A Phonetic Description of the Kawaiisu Language
Item Type text; Electronic Dissertation
Authors Thomas, Patrick Neal
Publisher The University of Arizona.
Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
Download date 05/10/2021 08:53:02
Link to Item http://hdl.handle.net/10150/631364 1
A PHONETIC DESCRIPTION OF THE KAWAIISU LANGUAGE
by
Patrick N. Thomas
______Copyright © Patrick N. Thomas 2018
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF LINGUISTICS
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
2018 2
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of the requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided that an accurate acknowledgement of the source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.
SIGNED: Patrick N. Thomas 4
ACKNOWLEDGEMENTS
First and foremost, my deepest gratitude goes to my wife, Megan Thomas, who has provided nothing but support over the last several years during my time as a graduate student at the University of Arizona. Megan, as well as our children (Claudia, Klaus, and Liam) have made the whole graduate experience much more bearable. Finishing a doctoral degree is a monumental task. It may be rewarding in the end, but it can also be stressful, frustrating, and overwhelming. It’s only due to the love and support provided to me by my family that I’ve been able to make it through graduate school with both a Ph.D. and my sanity. I’d also like to thank my parents for encouraging me from an early age to be inquisitive about the world around me, and to always learn new things. Without this kind of encouragement and support, I probably never would have decided to go to college in the first place, let alone attempt to earn an advanced degree! Special recognition goes to my father, Michael Thomas. Unfortunately, he passed away before my completion of the doctoral degree, but if it weren’t for all those trips to the library as a young kid, I’d probably be doing something completely different with my life now.
Although my graduate degrees come from the University of Arizona, I am indebted to my professors and other mentors from my undergraduate institution, the University of Oklahoma: Dr. Dylan Herrick, and Dr. Marcia Haag, who provided me with the inspiration to become a linguist, and taught me the foundational skills necessary to succeed (or at least not drown completely!) in a linguistics Ph.D. program. I am also grateful to Sophia Morren, the director of the Ronald E. McNair Postbaccalaureate Achievement Program at the University of Oklahoma, a program designed to help members of underrepresented groups gain access to graduate-level education. As a first generation college student, this program helped me tremendously, teaching me what to expect from a graduate program, as well as covering costs associated with graduate school applications.
My committee, Dr. Natasha Warner, Dr. Diana Archangeli, Dr. Ofelia Zepeda, and Dr. Amy Fountain, have graciously shared their time encouraging me, discussing my research with me, and leaving me tons of helpful feedback. If not for them, this research would not have been possible. I’d also like to thank Dr. Heidi Harley and Art Torrance for opening their home to my family when we first moved to Tucson. Moving an entire family (complete with spouse, kids, and pets) across the country is hard. They generously allowed us to stay in their home for a couple months while we searched for a new home in Arizona. Without them, relocating to Tucson for graduate school would have been impossible. I’d still be behind a cash register right now if not for them. I’d also like to mention Anna Norberg, my mentor during my previous life as an aspiring concert pianist, who believed in me when it seemed like not very many others did.
Finally, I am further indebted to my friends and colleagues at the University of Arizona as well as the University of Central Arkansas for their support, both social and intellectual. They include Gregory Anderson and Carl Eynatian, Jorge Muriel, Jaime Parchment, Catherine Huang, Brecht Welch, Ethan Dickinson, Ryan Walter Smith, Dr. Stacey Oberly, Dr. Andy Wedel, Dr. Adam Ussishkin, and many others. If I left you out, please don’t take it personally! I really haven’t forgotten you! It’s just that I’m at the bottom of the page here and I’m running out of 5
DEDICATION
To my family. 6
TABLE OF CONTENTS
LIST OF FIGURES ...... 9
LIST OF TABLES ...... 20
ABSTRACT ...... 25
CHAPTER 1: INTRODUCTION ...... 28
1.1 Importance of Phonetic Research ...... 28
1.2. The Genetic Affiliation of Kawaiisu...... 33
1.3. On the Kawaiisu Revitalization Effort ...... 42
1.4. Grammatical Structures of Kawaiisu ...... 46
1.4.1 Introduction ...... 46
1.4.2 Phonology ...... 47
1.4.3. Morphology...... 73
1.4.4 Basic Word Order ...... 76
1.5. Data and Methods ...... 78
1.6. Organization of the Dissertation ...... 83
CHAPTER 2: STRESSED VOWELS ...... 84
2.1 Introduction ...... 84
2.2 Methods...... 88
2.3 Results ...... 94
2.3.1 Duration Analyses ...... 94
2.3.2 Segmental Results ...... 98
2.4 Discussion ...... 117 7
CHAPTER 3: UNSTRESSED VOWELS ...... 120
3.1. Introduction ...... 120
3.2. Methods...... 122
3.3. Durational and Segmental Results ...... 123
3.3.1 Duration Analyses ...... 123
3.3.2 Segmental Results ...... 126
3.4. On Vowel Deletion ...... 144
3.4.1 Word-Internal Deletion ...... 144
3.4.2 Word-Final Deletion ...... 149
3.5. Discussion and Conclusion ...... 157
CHAPTER 4: ON THE ACOUSTIC CORRELATES OF STRESS IN KAWAIISU ...... 162
4.1 Introduction ...... 162
4.2. Methods...... 168
4.3. Results ...... 170
4.3.1 Duration ...... 170
4.3.2 Pitch ...... 173
4.3.3 Vowel Quality ...... 176
4.4. Discussion and Conclusion ...... 190
CHAPTER 5: ON CONSONANTS ...... 197
5.1 Introduction ...... 197
5.2 The Velar Obstruent ...... 198
5.3 Voice Onset Time ...... 207 8
5.4 Palatalization ...... 218
5.5. The Rhotic Consonant...... 234
5.6 The Glottal Stop ...... 242
5.6.1 Introduction ...... 242
5.6.2 Intervocalic Glottal Stops ...... 243
5.6.3 Pre-glottalized Sonorants ...... 248
5.7 Pre-aspirated Sonorants ...... 251
5.8 Conclusion and Discussion ...... 254
CHAPTER 6: CONCLUSION ...... 261
6.1 Introduction ...... 261
6.2 Summary of Previous Chapters ...... 262
6.3 Implications for the Community ...... 266
6.4 Questions for Future Research ...... 268
6.5 Conclusion ...... 269
APPENDIX: LIST OF TOKENS ...... 271
REFERENCES ...... 372 9
LIST OF FIGURES
CHAPTER ONE
FIGURE 1: Uto-Aztecan family tree…………………………………………………………….35
FIGURE 2: Internal classification of the Numic languages according to Freeze and Iannuci…..41
FIGURE 3: Hierarchical Representation of the Syllable………………………………………...58
CHAPTER TWO
FIGURE 1: Waveform and spectrogram of /keːvi/
/eː/, defined by onset of F2 and sudden drop in amplitude of F2………………………………..91
FIGURE 2: Waveform and spectrogram of /naˈvojo/
Kawaiisu speaker recorded in the 1980s. The selected portion indicates the stressed vowel
/o/………………………………………………………………………………………………...93
FIGURE 3: Comparison of long and short vowel durations in stressed position………………..95
FIGURE 4: F1 x F2 Vowel Plot from Speaker FC, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively………………………………………………………………………………………98
FIGURE 5: F1 x F3 Vowel Plot from Speaker FC, with F1 in Hz on the vertical axis, and F3 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..100 10
FIGURE 6: F1 x F2 Vowel Plot from Speaker AP, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..104
FIGURE 7: F1 x F3 Vowel Plot from Speaker AP, featuring F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..106
FIGURE 8: F1 x F2 vowel plot Vowel Plot from Speaker LG, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..109
FIGURE 9: F1 x F3 vowel plot from speaker LG featuring F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..111
FIGURE 10: F1 x F2 vowel plot Vowel Plot from Speaker RB featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..113
FIGURE 11: F1 x F3 vowel plot from speaker RB featuring F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by 11
squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..115
CHAPTER THREE
FIGURE 1: Comparison of long and short vowel durations in unstressed position, averaged over all speakers and tokens………………………………………………………………………….124
FIGURE 2: F1 x F2 Vowel Plot from Speaker FC, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..127
FIGURE 3: F1 x F3 Vowel Plot from Speaker FC, with F1 in Hz on the vertical axis, and F3 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………128
FIGURE 4: F1 x F2 Vowel Plot from Speaker AP, with F1 in Hz on the vertical axis, and F2 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………131
FIGURE 5: F1 x F3 Vowel Plot from Speaker AP, with F1 in Hz on the vertical axis, and F3 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by 12
squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..132
FIGURE 6: F1 x F2 Vowel Plot from Speaker LG, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………135
FIGURE 7: F1 x F3 Vowel Plot from Speaker LG, with F1 in Hz on the vertical axis, and F3 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..136
FIGURE 8: F1 x F2 Vowel Plot from Speaker RB, with F1 in Hz on the vertical axis, and F2 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………139
FIGURE 9: F1 x F3 Vowel Plot from Speaker RB, with F1 in Hz on the vertical axis, and F3 in
Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively……………………………………………………………………………………..140
FIGURE 10: Waveform and spectrogram of /tsiɣüˈpizi/
FIGURE 11: Waveform and spectrogram of /tsiɣüˈpizi/
is placed based on the burst of the /ɣ/…………………………………………………………..146
FIGURE 12: Waveform and spectrogram of /seːˈɣidü/
AP, with presence of word-final vowel………………………………………………………...150
FIGURE 13: Waveform and spectrogram of /seːˈɣidü/
AP, with word-final vowel deletion…………………………………………………………….151
CHAPTER FOUR
FIGURE 1: Comparison of duration between stressed and unstressed vowels, averaged across all vowels for all speakers. Phonemically short vowels appear in the left panel, phonemically long vowels in the right. Unstressed vowels are represented by blue bars, and stressed vowels by green…………………………………………………………………………………………….170
FIGURE 2: Comparison of pitch between stressed and unstressed vowels. Pitch values are in Hz.
Short vowels are shown in the left panel, long vowels in the right panel. Unstressed vowels are represented by blue bars, while stressed vowels are represented by green bars. Speakers are identified by initials across the bottom of the chart…………………………………………….174
FIGURE 3: F1 x F2 plot comparing the stressed and unstressed vowels of speaker FC. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….177
FIGURE 4: F1 x F3 plot comparing the stressed and unstressed vowels of speaker FC. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….178 14
FIGURE 5: F1 x F2 plot comparing the stressed and unstressed vowels of speaker AP. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….181
FIGURE 6: F1 x F3 plot comparing the stressed and unstressed vowels of speaker AP. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….182
FIGURE 7: F1 x F2 plot comparing the stressed and unstressed vowels of speaker LG. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….184
FIGURE 8: F1 x F3 plot comparing the stressed and unstressed vowels of speaker LG. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….185
FIGURE 9: F1 x F2 plot comparing the stressed and unstressed vowels of speaker RB. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….187
FIGURE 10: F1 x F3 plot comparing the stressed and unstressed vowels of speaker RB. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red…………………………………………………………….188
CHAPTER FIVE
FIGURE 1: Spectrogram and waveform of /tüˈɣahni/
The burst for /ɣ/ is highlighted in the spectrogram with arrows, indicating a clear stop 15
articulation and release for the /ɣ/ phoneme……………………………………………………200
FIGURE 2: Spectrogram and waveform of /seːˈɣidü/
LG. The release burst highlighted in the spectrogram (with arrows) indicates clear stop articulation and release of the word-medial /ɣ/ phoneme………………………………………201
FIGURE 3: Spectrogram and waveform of /seːˈɣidü/
FIGURE 4: Spectrogram and waveform of /paɣiˈkweːdü/
FIGURE 5: Spectrogram and waveform of /tuɣuˈkwidü/
The presence of formants throughout the production of this /ɣ/ indicates a likely approximant articulation……………………………………………………………………………………...204
FIGURE 6: Spectrogram and waveform of /joziˈküdü/
FIGURE 7: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors 16
shown are consonant phoneme, word position, stress level, and following vowel length. The data have been averaged across all four speakers……………………………………………………212
FIGURE 8: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme and word position. The data have been averaged across all four speakers…………………………………………………………………………………………213
FIGURE 9: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme and stress level. The data have been averaged across all four speakers…………………………………………………………………………………………214
FIGURE 10: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme and following phonemic vowel length. The data have been averaged across all four speakers……………………………………………………………….215
FIGURE 11: Waveform and spectrogram of /paˈtsaːtü/
4500Hz. The range shown in the spectrogram is 0 – 15kHz to show high frequency frication noise clearly…………………………………………………………………………………….220
FIGURE 12: Waveform and spectrogram of /patsaˈriːdü/
FIGURE 13: Waveform and spectrogram of /noɣotseˈʔedü/
FIGURE 14: Waveform and spectrogram of /aˈnisi/
FIGURE 15: Waveform and spectrogram of /wiˈhitsi/
LG. The fricative portion of the affricate /ts/ has the highest concentration of acoustic energystarting from approximately 2500Hz. The range shown in the spectrogram is 0-
15kHz…………………………………………………………………………………………...224
FIGURE 16: Waveform and spectrogram of /weːsaˈɣidü/
FIGURE 17: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ preceded by a front vowel, and followed by a consonant (in the /-vistü/ suffix)………………226
FIGURE 18: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ in the environment between two front vowels………………………………………………….227
FIGURE 19: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ in the environment between a front vowel (on the left) and a back vowel (on the right)………228
FIGURE 20: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ in the environment between two back vowels………………………………………………….229
FIGURE 21: Token count and distribution of acoustic energy for word-initial tokens of /s/, /ts/ immediately after a back vowel and immediately before a front vowel………………………..230
FIGURE 22: Token count and distribution of acoustic energy for word-initial tokens of /s/, /ts/ followed immediately by front vowels…………………………………………………………231 18
FIGURE 23: Token count and distribution of acoustic energy for word-initial tokens of /s/, /ts/ followed immediately by back vowels…………………………………………………………232
FIGURE 24: Waveform and spectrogram of /woʔoˈravi/
FIGURE 25: Waveform and spectrogram of /hoˈrodü/
AP. Contacts between the tongue and upper surface of the vocal tract are marked with arrows.
The dip in F3, indicating rhoticity, has been traced. The following rise in F3 has also been traced……………………………………………………………………………………………238
FIGURE 26: Waveform and spectrogram of /paʔatoˈɣorü/
FIGURE 27: Waveform and spectrogram of /püraˈvüni/
‘arm’ produced by speaker
LG, articulated with a flap. A single closure during the rhotic consonant has been indicated with an arrow. A slight F3 dip is visible……………………………………………………………..240
FIGURE 28: Waveform and spectrogram of of /püraˈvüni/
‘arm’ produced by speaker LG, articulated with a trill. Two contacts during the rhotic consonant have been indicated with an arrow…………………………………………………………………………241
FIGURE 29: Waveform and spectrogram of /jeˈʔedü/
Individual glottal pulses can be seen in the waveform and the spectrogram, indicating creaky voice. The final vowel has been deleted………………………………………………………..245
FIGURE 30: Waveform and spectrogram of /kotsoɣoˈʔadü/
by speaker LG. Glottal pulses can be seen in the spectrogram and waveform during the portion marked [ʔ]………………………………………………………………………………………246
FIGURE 31: Waveform and spectrogram of /viˈnoʔo/
The glottal stop appears as a brief decrease in amplitude………………………………………247
FIGURE 32: Waveform and spectrogram of /taʔniˈpüzi/
AP showing evidence of creaky voice………………………………………………………….249
FIGURE 33: Waveform and spectrogram of /taʔniˈpüzi/
LG without creaky voice………………………………………………………………………..250
FIGURE 34: Waveform and spectrogram of /pohˈnija/
LG. Glottal pulses are visible to indicate the presence of voicing, visible as vertical blue lines in the waveform…………………………………………………………………………………...252
FIGURE 35: Waveform and spectrogram of /tahˈmana/
LIST OF TABLES CHAPTER ONE
TABLE 1: The Consonants of Kawaiisu (Zigmond et al. 1990:5)………………………………47
TABLE 2: Kawaiisu Vowel Inventory Proposed by Zigmond et al. (1990)…………………….48
TABLE 3: Kawaiisu Consonant Inventory proposed by Klein (1959, 2002)……………………49
TABLE 4: Kawaiisu Vowel Inventory proposed by Klein (1959, 2002)………………………..49
TABLE 5: Kawaiisu orthography and IPA equivalents…………………………………………56
CHAPTER TWO
TABLE 1: Kawaiisu Phonological Vowels……………………………………………………...87
TABLE 2: Number of stressed vowel tokens available for analysis for each speaker and vowel phoneme (long and short vowels listed separately). This represents the number of each stressed vowel in the phonological environments described for inclusion……………………………….94
TABLE 3: F-ratios and significance levels comparing long and short vowel quality, for each vowel for speaker FC, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold………………………………………………………………….101
TABLE 4: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for FC. Significance is marked as indicated, and appears in bold……………………………………………………………………………………………..102
TABLE 5: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker AP, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold…………………………………………………….107 21
TABLE 6: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker AP. Significance is marked as indicated, and appears in bold………………………………………………………………………………….108
TABLE 7: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker LG, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold………………………………………………………111
TABLE 8: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker LG. Significance is marked as indicated, and appears in bold………………………………………………………………………………….112
TABLE 9: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker RB, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold………………………………………………………116
TABLE 10: F-ratios and significance levels comparing differences in timepoint (20%, 50%,
80%) on F1-F3 for each vowel phoneme for speaker RB. Significance is marked as indicated, and appears in bold……………………………………………………………………………..117
CHAPTER THREE
TABLE 1: Number of unstressed vowel tokens available for analysis for each speaker and vowel phoneme (long and short vowels listed separately). This represents the number of each unstressed vowel in the phonological environments described for inclusion…………………..124 22
TABLE 2: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker FC, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold…………………………………………………….129
TABLE 3: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker FC. Significance is marked as indicated, and appears in bold………………………………………………………………………………….130
TABLE 4: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker AP, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold…...………………………………………………….133
TABLE 5: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker AP. Significance is marked as indicated, and appears in bold………………………………………………………………………………….134
TABLE 6: F-ratios and significance levels comparing long and short vowel quality, for each of
LG’s unstressed vowels, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold……………………………………………………………….137
TABLE 7: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker LG. Significance is marked as indicated, and appears in bold………………………………………………………………………………….138
TABLE 8: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker RB, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold...…………………………………………………….141 23
TABLE 9: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker RB. Significance is marked as indicated, and appears in bold………………………………………………………………………………….142
CHAPTER FOUR
TABLE 1: F-ratios and significance levels comparing the effect of stress on vowel quality (F1,
F2, and F3) for each vowel phoneme for speaker FC. Significance is marked as indicated, and appears in bold………………………………………………………………………………….179
TABLE 2: F-ratios and significance levels comparing the effect of stress on vowel quality (F1,
F2, and F3) for each vowel phoneme for speaker AP. Significance is marked as indicated, and appears in bold………………………………………………………………………………….183
TABLE 3: F-ratios and significance levels comparing the effect of stress on vowel quality (F1,
F2, and F3) for each vowel phoneme for speaker LG. Significance is marked as indicated, and appears in bold………………………………………………………………………………….186
TABLE 4: F-ratios and significance levels comparing the effect of stress on vowel quality (F1,
F2, and F3) for each vowel phoneme for speaker RB. Significance is marked as indicated, and appears in bold………………………………………………………………………………….189
TABLE 5: Summary of differences in vowel quality between stressed and unstressed vowels for each speaker…………………………………………………………………………………….192
CHAPTER FIVE
TABLE 1: Number of stop consonant tokens available for VOT analysis. Table is organized by 24
speaker on the vertical axis, and by phoneme on the horizontal axis. Individual cells show the total token counts for each stop phoneme for each speaker, including token counts for specific conditions……………………………………………………………………………………….210 25
ABSTRACT This dissertation represents the first acoustic phonetic description of Kawaiisu, an endangered language of the Uto-Aztecan family spoken in the area near Tehachapi, California.
The Kawaiisu language is understudied and underdocumented. Although an impressive grammar
(with a lexicon and texts) has been published (Zigmond et al. 1990), discussion of the sound system is relegated to only a few pages. The present dissertation expands on the description of
Kawaiisu phonetics and phonology contained in Zigmond et al. (1990), finding support for many of the claims made in their grammar. In addition, I show that the phonetic system of Kawaiisu is characterized by rich variation, examining acoustic evidence to illustrate the different ways that
Kawaiisu sounds can be articulated.
The analyses presented in this dissertation come from recordings made by Sheldon Klein in the 1950s and 1980s (Klein 1958, 1981-1983). Digital versions of these recordings are hosted online at the California Language Archive, and numerous types of data are represented therein: wordlist elicitation, sentence elicitation, traditional narratives, monologues, and conversations in
Kawaiisu are all included in the archival data. In this dissertation, I examine words produced in isolation during elicitation sessions exclusively. These data come from four speakers: one male speaker recorded in the 1950s, and three female speakers recorded in the 1980s.
From these data, I examine both the consonant and vowel systems of Kawaiisu.
Considering vowels, evidence is presented which supports Zigmond et al.’s (1990) claim of a six-vowel system with phonemic vowel length contrast. I also find that an increase in pitch is the most robust acoustic cue to Kawaiisu stress, and that unstressed vowels often lie more toward the periphery of the vowel space than stressed vowels. Additionally, the pervasive phenomenon of 26
word-final vowel deletion in Kawaiisu is examined, claimed by Zigmond et al. (1990) to target verbs specifically. Although word-final vowel deletion is found to be a common process in
Kawaiisu, no evidence is found in the available data linking this process to a specific lexical category, and some suggestions are offered for why Zigmond et al. (1990) may have arrived at this conclusion.
Concerning consonants, this dissertation provides support for some of the phonological alternations mentioned by Zigmond et al. (1990), including the palatalization of coronal sibilants preceded by front vowels. I also show that many consonants have differing articulations in apparent free variation: the rhotic phoneme of Kawaiisu is variously realized as a flap or trill, while the consonant commonly labeled as /g/ (especially by Zigmond et al. (1990), and Klein
(2002)) may often be produced with fricative or approximant articulations, rather than as a stop.
Additionally, I show that the glottal stop (both intervocalically and in some cases pre- consonantally) is associated with the presence of creaky voice. Finally, I discuss voice onset time of voiceless stops, finding that VOT is often greater when the following vowel is phonemically long.
This research can be useful to both the community of linguists and to the Kawaiisu community. For linguists, this work presents a phonetic description of a language whose sound system was previously not adequately described. Variation in articulation has been noted where acoustic evidence for such variability was found, and several cross-linguistically rare patterns and structures have been indicated (for example, peripherality of unstressed over stressed vowels, the relationship between phonemic vowel length and VOT of stops, etc. ). For community members, the information presented in this document can be used to help create 27
pedagogical materials for purposes of language revitalization. Finally, this research highlights the importance of working with existing archival data. Archival language data is available through a number of sources, and in many cases, it may be all that remains of a once widely spoken language. This dissertation shows that important descriptive and documentary work can still be done, both to the benefit of the academic community, and to indigenous speech communities. 28
CHAPTER 1: INTRODUCTION
1.1 Importance of Phonetic Research The main goal of this research is to provide a detailed analysis of the phonetic structures of the Kawaiisu language, spoken in the area of Kern County, California, in and around the cities of Bakersfield and Tehachapi. This language is sometimes referred to as Nuwa, often by the speakers themselves (Kawaiisu Language & Cultural Center 2018). Despite the existence of freely available audio data, the phonetics of Kawaiisu remain virtually undocumented. When studying a language, especially an endangered language, it is advisable to describe all facets of the language as thoroughly and accurately as reasonably possible. Some documentation of
Kawaiisu has already taken place. For example, a grammar of Kawaiisu, complete with a reasonably extensive dictionary, exists (Zigmond et al. 1990). While it is an impressive and very important work, with detailed descriptions and analyses of Kawaiisu syntax and morphology, the grammar does not contain much detailed phonetic description outside of a few paragraphs on pronunciation in a short chapter on phonology. This is not terribly surprising, since detailed acoustic analyses of the phonetics of a language are rare in the grammar genre, but omitting such information leaves a crucial gap in documentation. The research presented in this dissertation, then, is one additional step towards a more complete documentation of the Kawaiisu language.
Documentation and description of a language can serve many purposes, both practical and theoretical. Such research can be important to the speech community whose language is the focus of study, as the research can go hand in hand with the development of pedagogical materials for language revitalization. Linguistic research can serve to affirm the importance of a given language as a serious topic of study, generating a sense of pride among community 29
members where use of an indigenous language has been maligned or discouraged by a superstrate cultural group. Yamada (2011), for instance, discusses her work with Kari’nja, spoken in Guyana and Suriname, detailing methods of community-based language documentation in which pedagogical materials are developed as part of the actual documentation process, noting the community’s pride in language and culture generated by this process.
From a scientific perspective, linguistic description is useful because any language being documented may have linguistic structures or paradigms which can advance our knowledge of how language works. Sapir’s (1949) documentary work on Southern Paiute, closely related to
Kawaiisu, provides a key example of the ways in which language documentation can have implications for linguistic theory. In this work, Sapir provides an anecdote wherein he asks a speaker of Southern Paiute to break up the word [pa:βa] ‘at the water’ into its constituent syllables:
“Tony then syllabified: pa: , pause, pa… Tony was not “hearing” in terms of the actual
sounds (the voiced bilabial β was objectively very different from the initial stop) but in
terms of an etymological reconstruction: pa: “water” plus postposition *-pa “at.” The
slight pause which intervened after the stem was enough to divert Tony from the
phonetically proper form of the postposition to a theoretically real but actually
nonexistent form”
(Sapir 1949:48-49).
30
In this example, the Southern Paiute word [pa:βa] is, as noted by Sapir (1949), morphologically complex, consisting of a root [pa:], followed by a locative suffix [-βa]. The speaker, however, seems to have the intuition that the initial [p] and the medial [β] are somehow the same, despite a surface difference in pronunciation. In the case of Sapir’s (1949) Southern
Paiute consultant, the phoneme /p/ is apparently the psychological equivalent of both the surface
[p] and the surface [β] in the Southern Paiute form [pa:βa:] ‘at the water.’ Thus, Sapir’s (1949) documentary work on Southern Paiute provides some evidence in favor of the existence of the phoneme as a linguistic unit. This is simply one example of a case where language documentation has enriched linguistic theory, but as languages disappear, it is important to describe and document them using available, ethical means, for the benefit of the speech communities themselves, as well as for enrichment of the field of linguistics.
While we know much about certain grammatical structures of Kawaiisu thanks to
Zigmond et al. (1990), Booth (1979), and Klein (2002), and we know about its genetic affiliation thanks to Klein (1959) and Iannucci (1973), there has been to date no serious study of the phonetic structures of the language. This work represents an attempt to fill in some of the missing pieces.
It is doubly unfortunate that such a study has not been done until this point, as the
Kawaiisu language is in a severe state of endangerment. Green (2013) notes that, in recent years, there were only three people who speak the language fluently, having learned it as a first language, or as young children. However, the number of speakers with at least some knowledge of Kawaiisu, such as heritage language learners, including those who learned Kawaiisu as a second language later in life, is possibly greater. With such a small number of L1 speakers, a 31
large scale phonetic study such as this one may impose too much on the valuable time of the L1 speakers. Thus, this study uses archival data to examine the acoustic phonetic structures of
Kawaiisu. The use of archival data for phonetic documentation has been established in the literature, including Tucker’s (2013) work on Chemehuevi, a Numic language closely related to
Kawaiisu.
While it is a simple thing to say that a language is endangered, quantifying the level of endangerment is another matter. Fishman (1991) introduces the Graded Intergenerational
Disruption Scale (GIDS), which measures the intergenerational continuity and vitality of a language community. In this scale, a higher score represents a more severe state of endangerment. The scale itself ranges from 1-8. In Stage 7 of the GIDS, Fishman (1991:89) notes that speakers of the language “are a socially integrated and ethnolinguistically active population but they are beyond child-bearing age.” As noted by Green (2013), most, if not all L1
Kawaiisu speakers are above the age of 60, which satisfies at least the GIDS Stage 7 requirement that all speakers of the language be beyond child-bearing age. Depending on the number of L2 speakers and heritage language learners, this may place some limits on social integration and ethnolinguistic activity.
In Stage 8 of the GIDS, Fishman (1991:88) notes that “most vestigial users of [the language] are socially isolated old folks and [the language] needs to be re-assembled from their mouths and memories and taught to demographically unconcentrated adults.” That is, a community suffering from Stage 8 language attrition is likely to have only a handful of elderly speakers who may not use the language with very many people very often. This seems to be the case with Kawaiisu, at least in the case of L1 native Kawaiisu speakers. In fact, Southern Ute, a 32
closely related language, has been described as falling into Stage 8 on the GIDS by Oberly
(2008), who reports that about 40 speakers remain. In her dissertation, Oberly (2008) collects phonetic data from several of these 40 speakers in a phonetic documentation project similar to the one being undertaken here. If it is the case that Southern Ute is a Stage 8 language, with 40 living speakers who are accessible enough to collect data from, then Kawaiisu, with a much smaller number of L1 speakers, should also be considered a Stage 8 language based on
Fishman’s (1991) GIDS.
There is no Stage 9 in Fishman’s (1991) scale. However, Lewis and Simons (2010) expand on the original scale, introducing Stages 9 and 10 in their Extended Graded
Intergenerational Disruption Scale (EGIDS). Stage 9 encompasses those languages considered to be dormant, where perhaps the last fluent speaker has recently passed away, but the community maintains a sense of identity with their language and wishes to attempt language revitalization.
In Stage 10, there are no remaining speakers, and no interest in the community to attempt to identify with or revitalize the language. In addition to these stages, Lewis and Simons (2010) split Fishman’s (1991) original Stage 8 into two different stages: Stage 8a and Stage 8b. They note that Fishman’s (1991) Stage 8 is analogous to their Stage 8a. In Stage 8b, Lewis and Simons
(2010:112) note that “the only remaining speakers are among the grandparent or great grandparent generation, and are so few or so scattered that they have little opportunity to use the language with each other.” This certainly sounds like the situation faced by the Kawaiisu community, based on the figures given by Green (2013). In fact, Kawaiisu is listed in Ethnologue
(Simons and Fennig 2017) as a Stage 8b language in Lewis and Simons’s (2010) EGIDS. 33
Despite the current absence of a sizeable Kawaiisu-speaking population from which to draw native, L1 speakers for phonetic documentation, a considerable quantity of Kawaiisu audio data does exist, recorded by Sheldon Klein in the 1950s and 1980s, and hosted by the California
Language Archive. The data contained in these archives is freely available online, and forms the basis of the analyses and description presented in this dissertation. Given the dire state of
Kawaiisu language endangerment, examining archival data allows for a relatively broad range of speakers to be considered. In this dissertation, I examine data from four speakers. This work not only fills in a gap in the literature by providing a phonetic description where there was none, but also highlights the importance of work dealing with indigenous languages, especially those that exist in various states of endangerment. Furthermore, this work will provide information that could be useful to Kawaiisu community members attempting to learn the language, either now, or if the language becomes dormant in the future. Although this dissertation itself is not suitable as teaching material per se, it is possible for the findings of this work to be presented in a useful way for the community.
1.2. The Genetic Affiliation of Kawaiisu Kawaiisu belongs to the Numic branch of the Uto-Aztecan family, one of the largest and most widespread language families in the Americas, stretching geographically from the U.S. state of Idaho in the north to the country of El Salvador in the south (Mithun 1999).
Much work has been done involving Uto-Aztecan historical linguistics, including work on the genetic affiliation of individual languages, internal classification, and on the grammar and reconstruction of the parent language, Proto-Uto-Aztecan (see, for example, Voegelin et al. 34
1962, Bascom 1965, Miller 1967, Crapo 1970, Hill and Hill 1970, Langacker 1970, Iannucci
1973, Steele 1973, Steele 1975, Langacker 1976a, Langacker 1976b, Heath 1977, Munro 1977,
Heath 1978, Fowler 1983, Manaster-Ramer 1984, Miller 1984, Heath 1985, Manaster-Ramer
1993, Stubbs 2000, Miller et al. 2005, Hill 2008, Stubbs 2011, and many, many others). The general consensus among Uto-Aztecanists is that the parent language, Proto-Uto-Aztecan, was spoken around 5000 years ago (Mithun 1999), although Holman et al. (2011) use computational techniques to arrive at an estimate of roughly 4000 years before the present.
The exact location of the Proto-Uto-Aztecan urheimat has been the subject of some controversy. Romney (1957) and Fowler (1983) suggest that the original Proto-Uto-Aztecan homeland was located in Arizona and northern Mexico, while Miller (1983) discusses evidence of a Proto-Uto-Aztecan homeland between the Mayo and Sinaloa rivers in northern Mexico. Hill
(2001) suggests that the Uto-Aztecan family may have originated further south in central
Mexico.
While scholars may argue on whether the exact location of the Proto-Uto-Aztecan homeland was in Arizona and northern Mexico or further south in Mesoamerica, it is generally agreed upon that Uto-Aztecan is a highly diverse language family, not just in terms of ecology and geographic spread, but also in its internal divisions. As people spread across North America from the Proto-Uto-Aztecan homeland (wherever it may actually have been), their languages changed over the centuries, eventually resulting in several different linguistic branches of Uto-
Aztecan, and dozens of individual languages. A family tree diagram for the Uto-Aztecan language family is shown below in Figure 1, based roughly on information from Mithun (1999). 35
FIGURE 1: Uto-Aztecan family tree
In Figure 1, the leftmost node represents the Uto-Aztecan language family itself. The names on the far right of Figure 1 are the individual languages that comprise the Uto-Aztecan language family. There are some intermediate nodes between the parent node and the individual languages. These intermediate nodes represent the different branches of the Uto-Aztecan family:
Numic, Takic, Tepiman, Taracahitic, Corachol, and Aztecan. The Numic and Takic branches are 36
divided further into sub-branches: Numic is divided into Central Numic, Southern Numic, and
Western Numic, while Takic is divided into two sub-branches – Serran and Cupan.
While all the languages on the far right of Figure 1 can trace their lineage back to Proto-
Uto-Aztecan, the nodes intervening between the Uto-Aztecan node and the individual languages show intermediate steps on the way. Consider, for instance, the two languages at the top right of
Figure 1: Mono and Northern Paiute. In the tree diagram, these two languages are connected immediately to the node labeled Western Numic. The implication of this connection is that, while Mono and Northern Paiute ultimately can be traced back to Proto-Uto-Aztecan, these languages can also be traced back to a hypothetical Proto-Western Numic, spoken much more recently. Likewise, the next cluster of languages in the figure, Timbisha, Shoshoni, and
Comanche, can be traced back to a hypothetical Proto-Central Numic. Looking further still,
Kawaiisu, Southern Ute, Southern Paiute, and Chemehuevi can be traced back to a hypothetical
Proto-Southern Numic. These branches, Western Numic, Central Numic, and Southern Numic, are connected immediately to a node simply labeled Numic, implying that Proto-Western Numic,
Proto-Central Numic, and Proto-Southern Numic (and, by extension, all the languages under them in the hierarchy) can all be traced back to a hypothetical Proto-Numic language, which in turn, can be traced all the way back to Proto-Uto-Aztecan.
In reality, the Uto-Aztecan family tree is more complicated than Figure 1 can indicate.
Many of the intricacies included in Mithun’s (1999) discussion of the genetic relationships in the
Uto-Aztecan language family are not included here. For instance, Mithun (1999) breaks down further some of the individual languages listed in Figure 1 as terminal nodes, separating them into different dialects. Where I simply list Guarijío, she distinguishes between Upland Guarijío 37
and Lowland Guarijío. She includes fine distinctions between over a dozen varieties of Shoshoni, and several varieties of Nahuatl (which she lists as Aztec). Additionally, for some of the branches, Mithun (1999) includes more details regarding intermediate sub-branches. For instance, if Figure 1 were simply an indiscriminate copy of Mithun’s (1999) groupings, the
Taracahitic branch would be separated further into Tarahumaran (which would include
Tarahumara and Guarijío), Opatan (which would include Opata and the distinct dialects and varieties she lists), Cahitan (which would include Hiaki and Mayo), and Tubar. However, others have pointed out that some of these fine sub-groupings may reflect linguistic similarities that have arisen due to areal contact, and that, in some cases, evidence for close genetic relationships may be somewhat tenuous (as noted by Iannucci 1973, Shaul 1983, and Haugen 2008). As it is not the goal of this dissertation to argue for or against these groupings, some potential intermediate sub-branches have been omitted from Figure 1.
Another major grouping in the Uto-Aztecan family which is neither shown explicitly in
Figure 1 nor in Mithun’s (1999) detailed list of Uto-Aztecan branches and languages (although she discusses the split elsewhere in the text) is a split between Northern Uto-Aztecan and
Southern Uto-Aztecan. This split is argued for by Hill (2008), Manaster-Ramer (1993), Heath
(1985), and Fowler (1983). Others (see Silver and Miller 1997) argue that the evidence for such a split is somewhat more tenuous. If, however, such a split had taken place, it is generally agreed that the Northern Uto-Aztecan branch includes Hopi, Tübatulabal, and the Numic and Takic sub- branches, while the Southern Uto-Aztecan branch includes the Tepiman, Taracahitic, Corachol, and Aztecan sub-branches. In Figure 1, the Northern Uto-Aztecan group contains the languages 38
and branches that are above the Uto-Aztecan node on the page, while the Southern Uto-Aztecan group contains the ones listed lower in the figure.
Kawaiisu is a member of the Numic branch. The name of this branch comes from the word for “person” in the Numic languages: Mono /nəəhmə/, Northern Paiute /nəhmə/, Timbisha,
Comanche, Shoshoni, and Southern Paiute /nəmə/, and Kawaiisu /ˈnüwü/ (Iannucci 1973,
Zigmond et al. 1990). These languages are spoken in the Great Basin region of the western
United States, in areas of California, Nevada, Arizona, New Mexico, Utah, Colorado, Wyoming,
Idaho, and Oregon. One exception is Comanche, which is spoken in western Oklahoma and the
Texas panhandle.
As with the Proto-Uto-Aztecan homeland, the location of the Proto-Numic urheimat has also been the subject of some debate. As discussed by Iannucci (1973), Lamb (1958) suggests that the Proto-Numic homeland may have been in the Death Valley area roughly 2000 years ago.
One reason for this suggestion is that the greatest range of diversity within Numic languages appears in this area. Taylor (1961), however, counters this line of reasoning, comparing the density of Numic languages in this region to the density of Celtic languages in the British Isles.
That is, Taylor (1961) suspects that the convergence of Numic languages in the Death Valley area was due to the migration of Numic speaking peoples from elsewhere, similar to how the
Celtic speaking peoples migrated to the British Isles from elsewhere in mainland Europe.
Miller (1966) argues that, for Taylor’s (1961) proposal to have any validity, Numic speaking groups would have had to simultaneously move into the Death Valley area in a very careful, orchestrated effort involving several pit stops along the way where some groups got left behind while others continued forward to Death Valley. Such a situation would be absurd, and, 39
as Miller (1996:93) points out, “it is unlikely that the Basin saw such carefully planned migration until the days of Brigham Young.” Iannucci (1973) argues in favor of a Death Valley homeland for Proto-Numic, noting that much linguistic diversity exists in the area, but diversity decreases substantially across the span of the Great Basin. This suggests that Proto-Numic speaking people originally lived in the Death Valley area, with many groups remaining in the area, while a few groups branched off over the centuries, spreading throughout the Great Basin.
Although the traditional grouping of Numic into Central, Southern, and Western branches is generally accepted, an alternative proposal is offered by Iannucci (1973) and Freeze and
Iannucci (1979). These sources argue in favor of a bipartite split from Proto-Numic into two sub- branches: Western Numic and Eastern Numic. In this view, Western Numic still consists of
Mono and Northern Paiute, while Eastern Numic contains all other Numic languages. Further subdivisions of Eastern Numic are problematic according to Iannucci (1973), who attempts to classify these languages based on shared morphophonological innovations.
The Numic languages are somewhat notorious for a complex set of patterns of word- medial consonant mutation. These patterns are explained in more detail in Section 4.2.5 of this chapter. Briefly, medial consonants in Numic languages can appear as plain consonants, pre- nasalized, pre-aspirated, pre-glottalized, geminated, spirantized, or, in some cases lenited entirely. This tends to happen to morpheme-initial consonants in morphologically complex words. The exact form of the consonant which appears depends on the preceding morpheme.
This process is responsible for Sapir’s (1949) anecdote, discussed above, in which a Southern
Paiute speaker decomposes the word [pa:βa] ‘at the water’ into the separate syllables [pa:-]
‘water’, and [-pa], a locative suffix. The initial [p] in the suffix is, of course, not the same as the 40
spirant [β], but surfaces as such due to the influence of the preceding morpheme, which dictates that a following consonant must be spirantized. The general analysis is that each morpheme is lexically specified to have a certain effect on a following consonant, and has nothing to do with phonetic environment (see, for instance, Charney 1993 for a detailed discussion of this phenomenon in Comanche). The exact patterns which surface vary between the different Numic languages.
The reason Iannucci (1973) finds dividing Eastern Numic into smaller branches difficult is that some languages which are not traditionally agreed to be to be sisters share similar sets of consonant mutation patterns. For example, Iannucci’s (1973) analysis, based on apparent shared innovations regarding these consonant mutation patterns, may place Southern Paiute
(traditionally regarded as a Southern Numic language) closer to Shoshoni (traditionally regarded as a Central Numic language) than to the other languages which traditionally form the Southern
Numic branch. Iannucci (1973) admits, however, that this analysis is tentative and warns readers that it should be evaluated critically and cautiously. Years later, Freeze and Iannucci (1979) consider grammatical and lexical evidence in addition to morphophonological consonant mutation patterns, arguing in favor of an Eastern Numic branch whose subdivisions contain the traditionally accepted Central and Southern Numic branches. A tree diagram of the internal division of Numic, as given by Freeze and Iannucci (1979), is shown below in Figure 2. 41
FIGURE 2: Internal classification of the Numic languages according to Freeze and Iannuci (1979).
Regardless of whether one accepts the traditional grouping of Numic into three branches
(Central, Western, and Southern), or the Freeze and Iannucci (1979) analysis which suggests a primary division between Western Numic and Eastern Numic, which is further subdivided into the Central and Southern branches, both the traditional grouping and the one shown in Figure 2 place Kawaiisu as a member of the Southern Numic group, closely related to Southern Paiute,
Southern Ute, and Chemehuevi.
As it is not likely to simply be the case that the Proto-Southern Paiutes, the Proto-
Southern Utes, the Proto-Chemehuevi, and the Proto-Kawaiisu broke apart from each other all at once, the Southern Numic branch can be subdivided even further. Kawaiisu has long been regarded as being the most unlike the other Southern Numic languages. The exact nature of these differences is beyond the scope of this dissertation, but a discussion of these differences can be found in Zigmond et al. (1990), especially in their chapter on comparative and typological notes to the grammar. In fact, many sources list Southern Paiute, Southern Ute, and Chemehuevi together as a single language or dialect continuum. Goddard (1983), in a consensus classification of North American languages, lists these languages together simply as Ute, with a parenthetical 42
noting each of the different groups. Silver and Miller (1997:371) list these as “Ute-Southern
Paiute, including Chemehuevi.” Elzinga (1999) calls this grouping Colorado River Numic.
Zigmond et al. (1990) refer to this grouping as a dialect chain, and are explicit in noting that
Kawaiisu is not a member of this chain. Thus, Kawaiisu is classified as a Southern Numic language in the Numic branch of Uto-Aztecan, and is highly divergent from the other Southern
Numic languages with which it shares its closest genetic affiliation. This suggests that, from the
Proto-Southern Numic speaking people, the Proto-Kawaiisu split from the main group at an earlier date from other Southern Numic people, whose languages form a dialect continuum in the
American Southwest.
1.3. On the Kawaiisu Revitalization Effort Wherever Proto-Uto-Aztecan and Proto-Numic may have been spoken, the Kawaiisu language, in more recent times, has been spoken in the area around Kern County, California, in and around the area of the present-day cities of Bakersfield and Tehachapi. In the introduction to the Zigmond et al. (1990:xi) grammar, Pamela Munro writes that, as of 1990, the language was spoken by “only a few people.” Hinton (1994) places the number of L1 Kawaiisu speakers in a range from 8 – 20. Mithun (1999) notes that, as of 1999, Kawaiisu was spoken by fewer than a dozen. The exact numbers are not clear for any of these estimates; however, Green (2013) notes that, by 2013, there were only 3 fluent L1 speakers. Again, these figures may not include speakers who learned the language later in life, or heritage language learners currently studying the language. However, it must be noted that L2 Kawaiisu speakers and other heritage language learners are important in the continued transmission of the language, and even though some 43
figures may not count such individuals, they play a vital role in keeping the language from falling into dormancy.
The U.S. Census Bureau reports 60 individuals identifying as Kawaiisu in 2010 (U.S.
Census Bureau 2010). This figure does not include information as to how many of those people actually speak Kawaiisu, and, as Hinton (1994) notes, figures provided by the U.S. Census
Bureau are often unreliable when it comes to surveying indigenous American populations. A larger number is given by the Kawaiisu Language & Cultural Center (2018), who note that there are about 250 living individuals who identify as Kawaiisu in and around Kern County.
The Kawaiisu Language & Cultural Center maintains a website, which can be accessed at http://www.kawaiisu.org, which details a community-based revitalization project. According to the site, the Kawaiisu Language & Cultural Center became an active non-profit organization in
2007, including a 9-member board. However, it is likely the case that community interest in language revitalization existed before this date, as such programs cannot exist without a great deal of community interest in the first place.
The website itself is detailed, and offers information on language revitalization in general, providing advice for other communities who wish to engage in language revitalization, noting that each speech community has its own specific needs and goals, and that what may work for one community’s revitalization program may not be suitable for another. They encourage community members who can speak the language to speak it at home, to children, friends, and loved ones, stressing the importance of linguistic immersion and intergenerational transmission. They also discuss the use of digital media in language learning and revitalization practices, noting that many community members have devices they can use to create high-quality 44
audio recordings of the language which can be studied anytime, anywhere. Additionally, issues relating to funding are discussed, noting that, while community interest in revitalization is necessary for a revitalization project to work, it is not sufficient, and that each community must have well-structured, well-defined goals that can be clearly communicated to funding agencies.
Such funds, if awarded, can be used towards immersion camps and schools, language classes, the development and creation of pedagogical materials, and other things, depending on the community’s needs and amount of available funds (see, for instance, Hinton 2002 on methods of language revitalization and the ways in which grant funds can be used to enrich such programs).
Also included on the website is a practical grammar, which explains the rules of the language in a clear and concise way, without unnecessarily complicated terminology or theoretical discussions. The goal is to create a grammar that members of the Kawaiisu community can read, understand, and use to enrich their knowledge of the Kawaiisu language.
Currently, several chapters of the practical grammar are available to download (in .pdf format) from the Kawaiisu Language & Cultural Center website. These include an introduction discussing how and why this project was created, a short note explaining basic grammatical terminology (as such terminology may sometimes be necessary to adequately explain how to use a given word that a user finds in the dictionary, for example), as well as chapters on the sounds of the language, verb tense and aspect, agreement (between verbs, subjects, and objects), nouns, pronouns, locatives and other descriptive suffixes and postpositions, wh-questions, adverbs, and syntax. The Kawaiisu examples used in the practical grammar chapters are written in an orthography based on the Roman alphabet, developed for and by the community. Many of these chapters are accompanied by Kawaiisu vocabulary lists and lessons in both audio and video 45
formats, which can be freely downloaded. Such projects can be particularly useful for language revitalization programs because many reference grammars (such as the Zigmond et al. 1990
Kawaiisu grammar) are written for an audience of professional linguists, and full of highly technical linguistics jargon, rendering them inaccessible for practical use among community members, who are not likely to have advanced training in theoretical linguistics. Additionally, such grammars are often out of print and hard to find outside of a university library.
Aside from this, several other projects relating to Kawaiisu revitalization are discussed on the Kawaiisu Language & Cultural Center website. The “Owot Abigip Nuwa” project was designed to help Kawaiisu language teachers reach community members who may be interested in learning their heritage language, and to develop additional lessons and pedagogical materials in addition to the practical grammar. Another project involved interviewing fluent speakers and recording conversations and stories in the language. Approximately 60 hours of audio were recorded as part of this project, which was funded by the National Endowment for the
Humanities and the National Science Foundation. Additional stories were video-recorded with
English subtitles, as part of a separate project funded by the Alliance for California Traditional
Arts and the Native Cultures Fund. The Kawaiisu Language & Cultural Center also worked on projects with the goal of developing Master-Apprentice programs designed to immerse families and individual language learners in the Kawaiisu language. One such project was funded by the
Sociological Initiatives Foundation, while other Kawaiisu Master-Apprentice programs are sponsored by the Advocates for Indigenous California Language Survival. Additionally, several
DVDs and CDs were developed, funded by the Alliance for California Traditional Arts, the
Native Cultures Fund, and the Ringing Rocks Foundation, designed to teach community 46
members about traditional plants used for both food and medicine. The Kawaiisu Language &
Cultural Center has also been involved in workshops designed for neighboring indigenous communities to create digital media to teach members of those communities about their own cultures and languages.
Additionally, the Kawaiisu Language & Cultural Center website hosts a calendar which provides information about ongoing language and cultural events for community members in the area. Past events have included community outreach events at parks and museums local to the
Kern County region, as well as an ongoing series of Kawaiisu language classes in the town of
Tehachapi, California. As of this writing, the most recent event listed on the calendar was Go
Native Day, on September 1, 2018, an event held at a local park in Tehachapi. This event was free to the public, and featured cultural demonstrations by members of the Kawaiisu community, and other nearby tribes (Kawaiisu Language & Cultural Center 2018, Hammond 2018).
1.4. Grammatical Structures of Kawaiisu 1.4.1 Introduction
In this section, I provide a basic overview of some aspects of Kawaiisu grammar. This is not intended to be an exhaustive description of the grammatical structures of Kawaiisu. Rather, this serves as a simple introduction to those aspects of the grammar which are referred to elsewhere in this dissertation. As this dissertation focuses primarily on the sounds of the
Kawaiisu language, I begin with a discussion of Kawaiisu phonology, introducing the phoneme inventories proposed by Klein (1959, 2002), and Zigmond et al. (1990). I also discuss the claims made by Thomas (2017) on the Kawaiisu vowel inventory. These claims are examined in closer 47
detail in Chapters 2 and 3 of this dissertation. Additionally, I discuss the phonology of Kawaiisu stress and some basic phonological alternations. The situation regarding Numic consonant mutations, discussed above briefly in Section 2 of this chapter, will be explored in more detail.
Aside from issues surrounding phonology, I provide brief discussions of Kawaiisu morphology, which, in some cases, interacts with the phonology affecting the location of stress.
Additionally, the basic word order of Kawaiisu is discussed in this section. Unless otherwise noted, grammatical information and examples in this section are taken from Zigmond et al.
(1990).
1.4.2 Phonology
1.4.2.1. Phoneme Inventory
The consonants of Kawaiisu are shown below in Table 1, taken from Zigmond et al.
(1990:5). The symbols which appear in Table 1 are the symbols originally used by Zigmond et al. (1990), to avoid multiplying the number of transcription systems used to describe the language. Those which differ from standard IPA usage are discussed. When two symbols appear in the same cell, the left symbol is voiceless and the right symbol is voiced.
TABLE 1: The Consonants of Kawaiisu (Zigmond et al. 1990:5). Labial Coronal Palatal Velar Labialized Glottal Velar Stops p b t d k kw ʔ Fricatives v s z š ž g gw h hw Affricates c č Nasals m n ŋ Rhotics r Laterals l Glides w y
48
There are a few symbols in Table 1 which are used somewhat differently from the standard IPA usage. First, the symbol /v/ is listed under a general labial column, rather than a more specific labiodental column, as this phoneme varies between labiodental [v] and bilabial [β] according to Zigmond et al. (1990). The symbol /c/ represents the affricate /t͡ s/, while the /č/ symbol represents the affricate / ͡tʃ/. Similarly, the /š/ and /ž/ of Zigmond et al. (1990) represent, respectively, IPA /ʃ/ and /ʒ/. The sound which Zigmond et al. (1990) identify with the /g/ symbol is actually a voiced velar fricative /ɣ/. Actual voiced velar stops do not occur in Kawaiisu outside of loanwords. Symbols which appear in Table 1 as digraphs actually represent single phonemes, so it may be more accurate to transcribe /kw/, /gw/, and /hw/ as /k w/, /ɣw/, and /h w/ (or perhaps
/hw/, either a labialized glottal fricative, or a preaspirated labiovelar glide depending on the analysis), respectively. Finally, the /r/ which appears in the coronal rhotic cell varies freely between the flap [ɾ] and the trill [r], while /y/ is the palatal glide /j/. The vowel inventory proposed by Zigmond et al. (1990) is shown below in Table 2.
TABLE 2: Kawaiisu Vowel Inventory Proposed by Zigmond et al. (1990). Front Back Unrounded Rounded High i ɨ u Mid e o Low a
In Table 2, the vowels presented by Zigmond et al. (1990) have their standard phonetic values, except for the vowel they represent with the symbol /ɨ/. In the IPA, the /ɨ/ symbol is used to denote a high unrounded central vowel, but it is used by Zigmond et al. (1990) to denote a high unrounded back vowel, whose standard IPA symbol is /ɯ/. They are quite explicit in noting this, writing that this sound is the “high, back, unrounded [ɯ]” (Zigmond et al. 1990:5). 49
Klein’s (1959, 2002) perception of the Kawaiisu phonemic inventory is slightly different, shown below in Tables 3 and 4. As above, I present consonants first, followed by vowels.
Adhering to the IPA standard, when two symbols appear in the same cell, the one on the left is voiceless, while the one on the right is voiced. The symbols appearing below are those used by
Klein (1959, 2002). Those symbols which differ from standard IPA usage will be discussed separately, followed by a discussion of how Klein’s (1959, 2002) inventories presented below differ from the inventories presented above, proposed by Zigmond et al. (1990).
TABLE 3: Kawaiisu Consonant Inventory proposed by Klein (1959, 2002) Bilabial Alveolar Velar Labiovelar Glottal Stops p b t d k g kw gw ʔ Fricatives s z h Affricates c dz Nasals m n Laterals l Glides j w
TABLE 4: Kawaiisu Vowel Inventory proposed by Klein (1959, 2002) Front Central Back High i y u Low e a o
Many of the symbols used by Klein (1959, 2002) do not differ significantly from their standard IPA usage. Like Zigmond et al. (1990), Klein (1959, 2002) treats the labiovelar stops as unit phonemes. Also as in Table 1, /c/ here represents the voiceless coronal affricate /ts/. The palatal glide /j/ is listed under the alveolar column, which may be a typographical error, or perhaps they simply did not wish to add an additional column just for one symbol. In Table 4,
Klein (1959, 2002) makes a curious choice in using the symbol /y/ to stand for what he considers 50
a high central vowel 1, for which the standard IPA symbol is /ɨ/. Interestingly, Klein’s fieldwork on Kawaiisu was conducted at roughly the same time as Lamb’s (1957) dissertation on the grammar of Mono, a closely related Numic language, which uses /y/ to represent the same vowel. Further, Pullum and Ladusaw (1996) note that the /y/ symbol is commonly used by
Slavicists to denote a high central, or high back unrounded vowel. It is clear, however, that the
/y/ used in Klein (1959, 2002) is intended to be understood as a high central vowel, and not a high rounded front vowel, as standard IPA usage would suggest.
Thomas (2017) examines acoustic data, concluding that the vowel represented by
Zigmond et al. (1990) as /ɨ/ and by Klein (1959, 2002) as /y/ is more accurately described as a mid-high central vowel [ɘ]. Thomas (2017) arrives at this conclusion by comparing the first and second formants of this vowel to those of the surrounding peripheral vowels /i/, /e/, /u/, and /o/, finding that the vowel occupies a space roughly between the front vowels (/i/, /e/) and the back vowels (/u/, /o/), and that the height tends to fall below that of /i/, but above that of /o/.
Setting aside the acoustic signatures of these sounds, there are a few key differences between the inventories presented by Zigmond et al. (1990) and those presented by Klein (1959,
2002); perhaps most obviously, the phonemic inventory proposed by Zigmond et al. (1990) is fuller than Klein’s (1959) phonemic inventory. Among the labial consonants, the /v/ present in the Zigmond et al. (1990) inventory is absent from the Klein (1959, 2002) inventory. Klein
(2002) argues that the bilabial stop [b] is in complementary distribution with a labial spirant
(possibly [β] or [v]), suggesting that the spirantized form exists as an allophone of the /b/
1 Actually, Klein (2002) refers to this as a high mid vowel, but it is clear that he uses the term ‘mid’ to refer to vowels which are intermediary along the front-back dimension, rather than as a term referencing vowel height.
51
phoneme, rather than as a distinct phoneme in its own right. While there are instances in which the labial stop alternates with the labial spirant, the distribution is not necessarily complementary, a fact implied by Zigmond et al. (1990), relating to developments regarding the system of Numic consonant gradation discussed later in this section.
Also present in the Zigmond et al. (1990) inventory but missing from the Klein (1959,
2002) inventory, is the rhotic /r/. This omission is not mentioned specifically by Klein (2002), but it is implied that the relationship between [r] and [t] is similar to the phonological relationship between the bilabial stop and spirant, mentioned above. Thus, it may be the case that a distinct /r/ phoneme does not appear in any of the phonemic inventories presented in Klein
(1959, 2002) because he analyzes [r] as simply being an allophone of /t/.
While Zigmond et al. (1990) and Klein (1959, 2002) differ in whether they choose to analyze /r/ as being phonemic, both sources include a phonemic lateral /l/. This is an interesting choice, because [l] only occurs in loanwords, never in native Kawaiisu forms. Similarly, the velar nasal [ŋ] also only occurs in loanwords. Zigmond et al. (1990) acknowledge this, but include both of these sounds as Kawaiisu phonemes anyway due to their frequent appearance in common loanwords. That Klein (1959, 2002) includes /l/, despite its appearing only in loanwords, but omits /ŋ/ (which occurs in the same circumstances), is perplexing.
One final difference between these competing inventories concerns the different representations of the voiced velar consonants, both plain and labial. Zigmond et al. (1990) prefers to treat these as fricatives /ɣ/ and /ɣw/ (although the symbol /g/ is consistently used for a voiced velar fricative in their text), while Klein (2002) argues that these should be analyzed as 52
stops. In Chapter 5, I present acoustic evidence in favor of the Zigmond et al. (1990) assertion that these should not be treated as stops.
1.4.2.2. On Symbols
The discussion in Section 4.2.1 above touches on the issue of uniformity with regard to which symbols are used to stand for a given sound. For the majority of Kawaiisu phonemes,
Klein (1959, 2002), Zigmond et al. (1990), and Thomas (2017) are mostly in agreement on their symbolic representations. Some symbols appearing in these sources, as noted, are used in ways which differ from the IPA standard usage. Where this occurs, I have noted the differences and attempted to make the intended usages clear so that readers will be aware which symbols stand for which Kawaiisu sounds.
While these sources are mostly in agreement on usage of symbols, one phoneme where they are not in agreement is the vowel described by Zigmond et al. (1990) as being high, back, and unrounded. This is the vowel represented by Klein (1959, 2002) as /y/, by Zigmond et al.
(1990) as /ɨ/ or /ɯ/, and by Thomas (2017) as /ɘ/. Each of these sources uses a different symbol for this vowel. While the number of different symbols used by different sources for the same
Kawaiisu vowel can be confusing, each symbol has its advantages. The /ɨ/ symbol used by
Zigmond et al. (1990) is widely used in the literature to stand for this vowel in many related
Numic languages (see for instance Booth 1979, Hill 2008, and Toosarvandani 2010). However, this vowel is standardly used to indicate a high vowel, and Thomas’s (2017) results (as well as the results presented in Chapters 2 and 3 of this dissertation) indicate that this vowel may vary in height between high and mid. 53
The /ɯ/ symbol, also used by Zigmond et al. (1990), is used in the IPA to stand for a high, back, unrounded vowel. This is an attractive choice because, as argued by Thomas (2017), the vowel seems to pattern phonologically as such. However, the results of the acoustic analyses in both Thomas (2017) and in this dissertation indicate that the vowel is often further forward in the vowel space than the back vowels, and so employing a symbol which is usually used to indicate a back vowel may be misleading.
Klein’s (1959, 2002) choice of /y/ for this vowel may confuse readers who are used to the standard IPA usage of this symbol, expecting it to stand for a high, front, round vowel. As noted by Pullum and Ladusaw (1996), the /y/ symbol has been used by Slavicists in the past to stand for a vowel which can vary in height between high and mid, and in backness between central and back, but this usage is no longer common.
The choice presented in Thomas (2017), /ɘ/ is standardly used in the IPA to indicate a close-mid central unrounded vowel. Using this symbol highlights the phonetic centrality of this vowel in the Kawaiisu vowel space, as well as the fact that it tends to vary in phonetic height between high vowels like /i/ and /u/, and mid vowels like /e/ and /o/. From a phonological perspective, however, using this symbol ignores the fact that the vowel patterns phonologically as a high back vowel. Zigmond et al. (1990) use two symbols for this vowel, noting that the vowel phonetically may be considered [ɯ], but they prefer to use the symbol /ɨ/ in their text. If we accept the practice of using different symbols to refer to the same vowel, then perhaps it may be more accurate to say that this vowel phonetically may be considered [ɘ], but that it phonologically patterns as /ɯ/. 54
With these issues in mind, the question remains: which symbol is the best to use? While each of the choices above has a unique advantage, they all have drawbacks as well: some of the choices ignore phonetic patterns, and others ignore phonological patterns. With the /ɘ/ choice, we must be mindful of the fact that an extra symbol (that is, /ɯ/) may be necessary depending on whether the discussion is one of phonetics or of phonology. None of the above choices are quite perfectly suited for the job of symbolizing this particular Kawaiisu vowel phoneme. Thus, in this dissertation, I use the symbol /ü/. First and foremost, this is the symbol which appears in the practical grammar hosted online by the Kawaiisu Language & Cultural Center (2018).
Considering some of the drawbacks surrounding other symbols for this vowel, it is fitting to use a symbol which has already been adopted by the speakers themselves. Further, using the /ü/ symbol is not too far out of line from the way this symbol might be standardly used in IPA transcriptions. The tréma diacritic above this symbol is standardly used to indicate a degree of centralization. Thus, the symbol /ü/ can be used to stand for a vowel which is both high and back
(as it patterns phonologically), with the diacritic above the symbol used to indicate that the vowel lies closer to the center of the vowel space than a typical /u/ articulation. This symbol, then, has numerous advantages: using a symbol that is already used by the Kawaiisu themselves helps to make this work slightly more accessible to those it may benefit, and it also captures some of the phonetic and phonological peculiarities which make it difficult to symbolize. In this dissertation, the symbol /ü/ is always used for this Kawaiisu vowel phoneme, unless a different source is being directly quoted. In these cases, the symbol used by the original author or authors is presented.
55
1.4.2.3 Current Kawaiisu Orthography
The Kawaiisu have adopted their own orthographic conventions, which are separate from
(but ultimately similar to) the conventions used in the IPA transcriptions seen in this dissertation.
Wherever possible in this dissertation, when discussing Kawaiisu words, I provide the IPA transcription and an orthographic transcription, attempting to use the spelling system adopted by the Kawaiisu themselves. The finer details of the Kawaiisu orthography can be viewed on their website (Kawaiisu Language & Cultural Center 2018), but I have attempted to provide a table listing each phoneme of Kawaiisu, and its orthographic equivalent, below in Table 5. Because
Kawaiisu spelling is not standardized, some speakers may spell the same words in different ways. Therefore, the way I transcribe a Kawaiisu word orthographically may not be the same way that every Kawaiisu speaker will spell the same word. In general, some variation is allowed, but in errors in phonemic or orthographic transcription are my own. In the rest of this dissertation, phonemic transcriptions are given in slashes / /, phonetic transcriptions are given in square brackets [ ], and orthographic transcriptions are given in angled brackets < >. In the table below, where allophony results in a difference in orthographic transcription, the IPA transcription column includes the phoneme in slashes, followed by the allophone in square brackets. Symbols in this table are listed in alphabetical order. 56
TABLE 5: Kawaiisu orthography and IPA equivalents Consonants Vowels IPA Transcription Orthographic IPA Transcription Orthographic Transcription Transcription /b/ /a/ /ts/ [tʃ]
/p/
/r/ /s/ [ʃ]
In addition to the phones and their orthographic equivalents given in the table above, the
Kawaiisu orthography uses some other symbols whose phonetic or phonemic equivalents are not immediately obvious, which may result from processes of allophony not studied in this work.
These include the symbols <ä> and <ö>. Additionally, word-final vowels are sometimes written in parentheses, to highlight the fact that these vowels are optionally deleted or otherwise 57
devoiced by some speakers in many environments. This process of word-final vowel deletion will be discussed further in Chapter 3 of this dissertation.
1.4.2.4. Kawaiisu Stress and Prosodic Units
Before diving into a discussion of Kawaiisu stress, it is necessary to introduce the units of prosody that are relevant to stress assignment in this language: the syllable and the mora.
Hierarchical organization is a key characteristic of language. At practically all levels, linguistic units can be broken down into even smaller units, which can themselves be broken down further still. At the level of sentence structure and meaning, a sentence can be broken down into its syntactic constituents, and these phrases can be broken down into individual words, and these words can be broken up into their constituent morphemes.
Likewise, at the phonological level, words can be broken up based on their prosodic constituents. The syllable is a particularly salient prosodic structure which speakers often have strong intuitions about. Regarding the syllable, Haugen (1956:196) notes, “speakers of the language can utter them separately, dividing utterances into sequences that seem natural when pronounced alone.” For instance, a native speaker of American English who is asked about the syllables in a word like Alabama should have no difficulty breaking up the word into individual syllables such as [æ], [lə], [bæ], and [mə], although the speaker may not be able to articulate what unifies these chunks as cohesive units of prosody, and speakers may put syllable boundaries at slightly different locations in some instances.
Generally, the syllable is conceived as a sequence of one or more segments, containing at least a nucleus (usually a vowel, although some languages allow consonants as syllabic nuclei, 58
see Bell 1978), optionally preceded and followed by a consonant (or clusters of consonants)
(O’Connor and Trim 1953). The term onset is generally used for consonants within a syllable preceding the nucleus, and the term coda is used for those consonants which follow the nucleus
(Hockett 1955). The result of this configuration, as discussed by Whitney (1874) and Saussure
(1922) is that, within a syllable, the nucleus is loudest and most sonorous, with sonority rising into the nucleus from an initial consonant or consonants (if any), and falling from the nucleus into a final consonant or consonants (if any). Nuclei and codas are often analyzed as belonging to a subsyllabic unit known as the rhyme, leading to the following hierarchical representation à la
Fudge (1969), shown below in Figure 3 (in this figure and elsewhere, the σ symbol represents a syllable, C indicates a consonant, and V indicates a vowel).
FIGURE 3: Hierarchical Representation of the Syllable
Figure 3 exemplifies a typical syllable with a vowel nucleus, a consonantal onset, and a consonantal coda. In reality, not every syllable in every language will necessarily have an onset or a coda. Onsets are obligatory in some languages and optional in others, while codas are disallowed in some languages and optional in others. Languages are not likely to ban onsets or require codas (although see Breen and Pensalfini 1999 for a discussion of a language which apparently lacks consonantal onsets). 59
This general cross-linguistic preference for onsets over codas has some implications for the ways in which syllables tend to be parsed. Clements and Keyser (1983:38) discuss the Onset
First Principle, writing that, “given alternative syllable divisions, languages will select that which maximizes syllable-initial consonant sequences.” For example, a VCV string in a language is likely to be parsed as .V.CV., rather than .VC.V., even if .VC. is a legal syllable in the language, due to the tendency of languages to prefer syllabification of onsets over codas. Ito (1987) arrives at a similar conclusion, although it is stated in slightly different terms, requiring a CV sequence to be tautosyllabic. As with the Onset First Principle discussed by Clements and Keyser (1983), this results in a VCV sequence syllabified as .V.CV. In short, languages generally prefer consonants to be syllabified as onsets if possible, with codas only occurring in situations where syllabifying the consonant as an onset would violate other rules of syllable structure in the language. In the case of Kawaiisu, syllable structure tends to be fairly simple. According to
Zigmond et al. (1990), most Kawaiisu syllables are of CV type, with CVC syllables appearing rarely.
The mora is a subsyllabic unit which determines the weight of syllables. The mora, often denoted by the μ symbol, has been described by McCawley (1968:57) as “something of which a long syllable consists of two and a short syllable consists of one.” Hayes (1989:256) writes explicitly, “in languages with contrastive vowel length, long vowels have two moras, short vowels one.” In many languages, coda consonants are often analyzed as being moraic. This is language specific. For instance, Hayes (1989) notes that codas are moraic in Latin, but not in
Lardil (an indigenous language of Australia). According to the discussion of Kawaiisu phonology in Zigmond et al. (1990), coda consonants in Kawaiisu are exceedingly rare, so it is 60
not particularly relevant to stress placement whether or not coda consonants add to the mora count in Kawaiisu.
In a given language, stress can be assigned to a certain syllable, or to a certain mora. In
Kawaiisu, stress is predictable and regular. Zigmond et al. (1990) write that stress is carried by the penultimate mora of the Kawaiisu word. In many Kawaiisu words, the penultimate mora also falls on the penultimate syllable, but in words that end with a long vowel (which occur rarely), the penultimate mora falls on the final syllable, since long vowels, as noted above, are analyzed as having two morae. Compare the two forms from the dictionary portion of Zigmond et al.
(1990), shown below with an acute accent to mark stress and periods to mark syllable boundaries.
(1) Stress placement in Kawaiisu 2
a. Word-final long vowel b. No word-final long vowel
[.pa.jáa.]
One might think these data suggest that stress in Kawaiisu is variable and can occur either on the final syllable or the penultimate. However, if the relevant prosodic unit is taken to be the mora, as Zigmond et al. (1990) suggest, rather than the syllable, stress becomes predictable. Below, (2) shows the same words, including their moraic structure, making this generalization regarding stress assignment and mora count more explicit.
2 Words ending in long vowels are rare in Kawaiisu. In fact, no such words were noted in my analysis of the Klein recordings which are examined in this dissertation (Klein 1958, 1981-1983). There are only a few of these words in the Zigmond et al. (1990) dictionary, and they do not provide any specific examples regarding stress placement in such words. If their statements on stress placement are correct, however, then the final syllable of [pajáa] should be stressed. 61
(2) Moraic representation of Kawaiisu words
a. [.pa.jáa.]
σ σ σ σ σ
μ μ μ μ μ μ μ
p a j á p o p í ʒ i
As vowels in Kawaiisu are moraic, each vowel in (2) is connected to at least one mora.
Short vowels have only one mora, while long vowels have two. This being the case, the final long vowel in [.pa.jáa.]
Likewise, the stressed vowel in [.po:.pí.ʒi]
Some processes may slightly obscure this general pattern. For example, vowels in word- final position are optionally deleted by some speakers. In these cases, stress seems to appear on the word-final syllable.
(3) Stress and Final Vowel Deletion
(a). [.wi.ná.pi]
(b). [.wi.náp.]
The forms in (3) are both the same lexical item, with the difference being that the final vowel is deleted in (3b). Assuming that each vowel in (3a) is associated to a single mora, the generalization that stress falls on the penultimate mora still holds. Logically, there are three possible analyses for (3b), which has the surface appearance of word-final stress. First, one could simply argue that forms like (3b) are exceptions to the general pattern, but this is unsatisfying.
Second, based on the surface form alone, and the fact that codas are often moraic (as Hayes 1989 62
points out), one could argue that the word-final [p] in (3b) constitutes a moraic coda, and as such, the stress in (3b) still falls on the penultimate mora. Finally, one could argue that stress assignment is based on the underlying form, /winapi/. In this case, the final /i/ would normally be associated to a mora, and stress would fall on the preceding mora associated to /a/. However, due to the optional process of word-final vowel deletion, the /i/ does not appear in the phonetic form, but /a/, which is still associated to the penultimate mora in the underlying form, receives stress. A similar phenomenon occurs in Japanese, in which devoiced or deleted vowels are moraic and can be used to predict pitch accent (Hasegawa 1999). In either of the latter two arguments, stress is still carried by the penultimate mora, rather than being seen as an exceptional anomaly.
1.4.2.5. Phonological Alternations
In language, it is not the case that each phoneme is pronounced exactly the same way in each word or phonological environment. For example, in many dialects of English, the /l/ phoneme appears as the velarized [ɫ] in syllable-final position, and as [l] syllable-initially.
In a given language, many such phonological alternations exist, with different variants of a phoneme surfacing depending on phonological factors. Zigmond et al. (1990) identify two such phonological alternations in Kawaiisu that will prove to be relevant to this dissertation.
First, Zigmond et al. (1990) identify a pattern wherein coronal fricatives and affricates alternate with palatalized fricatives and affricates. Examine the forms below in (4). Symbols used in these examples are the same used by Zigmond et al. (1990).
63
(4) Kawaiisu Palatalization (Zigmond et al. 1990:7)
(4a) [ʔɨkɨ-ci] <ükütsi> ‘little hat’ (4d) [šivoroni-či]
hat-DIM basket-DIM
(4b) [pugu-zi]
dog-ABS owl-ABS
(4c) [hana-sɨbi]
who-IRR what-IRR
In these examples, the coronal allophones [c], [z], [s] surface in (4a-c), while their palatalized counterparts [č], [ž], [š] appear in (4d-f). The pattern, as noted by Zigmond et al.
(1990), is that the palatalized forms surface when the preceding vowel is a front vowel like [i].
Elsewhere, when the preceding vowel is not a front vowel like [i], the coronal forms surface.
Acoustic evidence illustrating this phenomenon is examined in Chapter 5 of this dissertation.
Aside from this pattern of palatalization, Zigmond et al. (1990) also discuss a pattern of vowel coalescence. This occurs when a vowel-initial clitic is attached to a vowel-final stem.
When these vowels are the same, the result is a long vowel. When these vowels are not the same, one might expect the result to be a sequence of two non-identical vowels. However, sequences of non-identical adjacent vowels are comparatively rare in Kawaiisu, and are usually avoided through vowel coalescence. The actual result of this concatenation, in the pronunciation which surfaces, is a long vowel which combines features of the two original, or underlying vowels.
Consider the data in (5) below, taken from Zigmond et al. (1990).
64
(5) Vowel Coalescence (Zigmond et al. 1990:11-12)
(a) /a=i/ ‰ [eː] /kaʔa-na=ina/ ‰ [kaʔaˈneːna]
eat-CMP =his ‘his eating’
(b) /ü=a/ ‰[aː] /pükeː-dü=aka/ ‰ [pükeːˈdaːka]
see-NMR =it ‘see it (that one)’
(c) /ü=i/ ‰ [iː] /pükeː-dü=ika/ ‰ [pükeːˈdiːka]
see-NMR =it ‘see it (this one)’
(d) /u=a/ ‰ [oː] /karü-ɣu=su=ana/ ‰ [karüɣuˈsoːna]
sit-DS =EMPH =his ‘while he (that one) sat’
(e) /u=i/ ‰ [iː] /karü-ɣu=su=ina/ ‰ [karüɣuˈʃiːna]
sit-DS =EMPH =his ‘while he (this one) sat’
(f) /a=u/ ‰ [uː] /hivi-taʔa=aka=umi/ ‰ [hivitaʔaˈkuːmi]
drink-INT =it=your: PL ‘you ( PL ) may drink it’
Thomas (2017) discusses these alternations in more detail, noting that these patterns adhere to Casali’s (1996) principles of Feature Preservation Preference and Position Preservation
Preference. Essentially, if either vowel in the sequence is non-high, then the result which surfaces is also non-high. Additionally, aside from height features, all other features of the second vowel are retained in the surface form. The analysis presented in Thomas (2017) has implications for the featural specification of the vowel represented by Zigmond et al. (1990) as
/ɨ/, and will be revisited in Chapter 2 in a more in depth discussion of the Kawaiisu vowel system.
65
1.4.2.6. Numic Consonant Mutations
No discussion of the phonology of a Numic language can be complete without mention of the complex system of consonant mutations which are pervasive in this branch of Uto-Aztecan.
This section begins with a general discussion of Numic consonant mutation, showing how widespread this phenomenon actually is in this language family, before moving into a brief discussion of consonant mutation in Comanche, with specific examples to show how this process works in a language where it is still somewhat predictable. Finally, we return to Kawaiisu, where, according to Zigmond et al. (1990), only vestiges of the consonant mutation system remain, the historic reflexes of which are either fossilized or otherwise no longer predictable.
In Numic languages, the consonants which surface (especially morpheme-initially, although see Voegelin et al. (1962) for discussion of how these processes may operate morpheme-internally) typically vary in a systematic way. Generally, Numic consonant mutations are analyzed as being predictable on a morphological basis, rather than on phonological environment (Iannucci 1973). That is, the consonant which surfaces at the beginning of a morpheme, in many Numic languages, is said to depend on a lexically specified feature of the preceding morpheme, and not on the phonological features of any adjacent segments.
For instance, Sapir’s (1931) work on Southern Paiute shows that, in intervocalic position, the only stops which surface are spirants, geminates, or prenasalized stops. What Sapir (1931) finds, specifically, is that certain morphemes always trigger the presence of a spirant, other morphemes cause a geminate to occur, and other morphemes are associated with a following prenasalized stop. In his analysis, Sapir (1931) thus divides Southern Paiute morphemes into three main types: spirantizing, geminating, and nasalizing. In this analysis, there is no way to tell 66
which of the three types a given Southern Paiute morpheme belongs to. In other words, this information is purely lexical, and speakers must simply know which effect a given morpheme has on a following consonant.
This analysis is discussed further by Iannucci (1973), who notes the Southern Paiute morphemes /ta-/ ‘foot’, /-šəu-/ ‘digit’, and /-paa-/ ‘at’ as belonging to the geminating, spirantizing, and nasalizing morpheme types, respectively. If a Southern Paiute speaker knows that these morphemes belong to these respective morpheme types, and knows that the next morpheme in a word begins with the /p/ phoneme, then the speaker can correctly select the geminate allophone [pp] following the geminating /ta-/ morpheme, the spirant allophone [β] following the spirantizing /-šəu-/ morpheme, or the pre-nasalized [ mb] allophone following the nasalizing /-paa-/ morpheme. There is nothing inherent about the segments in these morphemes or their individual phonological features which result in a following spirant, geminate, or pre- nasalized stop; rather, it is the case that a morpheme marked as a spirantizing morpheme will always be followed by a spirant, a morpheme marked as a geminating morpheme will always be followed by a geminate consonant, and a morpheme marked as nasalizing will always be followed by a nasalized consonant.
Similar processes occur in other Numic languages. The exact way these processes work differs from language to language, but it is always the case that the surface forms of morpheme- initial consonants depend on a lexically specified feature of a preceding morpheme. In Mono and
Northern Paiute, these processes cause consonants to appear either as spirants or geminates
(Iannucci 1973). In Chemehuevi, the operative consonantal processes are similar to Southern
Paiute, in that consonants surface as spirants, geminates, or are nasalized (Press 1979). In 67
Timbisha and many dialects of Shoshoni, consonants can be spirantized, geminated, nasalized, or pre-aspirated (Dayley 1989, Crum and Dayley 1993, Charney 1993), and in Comanche, morpheme-initial stops can be spirantized, aspirated, pre-aspirated, or may surface as plain stops
(Charney 1993). Thus, while the consonant mutation processes exist in some form across most of the Numic languages, the specific surface forms which can occur are language specific.
According to Zigmond et al. (1990), these processes no longer occur with predictable regularity in Kawaiisu. Comanche, however, has been well-studied and its consonant processes thoroughly described by Charney (1993). Therefore, let us turn to Comanche for some brief examples of Numic consonant mutation in action.
Scholars who study Numic languages often refer to these consonant mutations in terms of
‘final features’ (Iannucci 1973, Nichols 1974). This is a theory-neutral term, and, as Elzinga
(1999:13, fn. 3) writes, “does not presuppose an analysis based on distinctive feature theory.”
Recall that, in Southern Paiute, Sapir (1931) claims that speakers must know whether a specified morpheme belongs to the geminating, spirantizing, or nasalizing type. The final features may be thought of as the lexical specification which assigns morphemes as being one ‘type’ or the other.
In her work on Comanche, Charney (1993) marks final features with the following notation:
(6) Comanche Final Feature Notation (Charney 1993)
SYMBOL PROCESS DESCRIPTION
(a) /H/ pre-aspiration Causes a stop to be pre-aspirated
(b) /h/ aspiration Causes a stop to surface as a voiceless fricative
(c) /=/ fortition Causes a stop to surface as a plain stop
(d) unmarked spirantization Causes a stop to surface as a voiced fricative 68
In Comanche, according to Charney (1993), four consonantal processes are operative.
Consonants which appear word-medially and morpheme-initially may appear as pre-aspirated, as a voiceless fricative, a voiced fricative, or as a plain stop. 3 The symbols in the left column of (6) are used by Charney (1993) to symbolize Comanche final features, which indicate the specific
‘type’ of a given morpheme. Using this notation, a morpheme marked with /H/, for example, belongs to the pre-aspirating type, which causes a following stop to be pre-aspirated. Because this information must be lexically specified, or, in phonological terms, belongs to the underlying form, the final feature notation is placed in slashes. Some Comanche examples are shown below, followed by a brief discussion:
(7) Comanche Consonant Mutation Examples (Charney 1993)
UNDERLYING SURFACE GLOSS CONSONANTAL PROCESS(ES)
(a) /atɨka-paʔa =/ [arɨkavaʔa] ‘on the deer’ spirantization
(b) /kwasuʔu=-paʔa=/ [kwasuʔupaʔa]‘on the dress’ fortition
(c) /saapɨh-paʔa=/ [saapɨ̥ ɸaʔa] ‘on the stomach’ aspiration
(d) /wɨH-pitɨ/ [wɨhpitɨ] ‘to be near something’ pre-aspiration
The items in (7) display various surface manifestations of the /p/ phoneme in Comanche, based on the consonant mutation process. Items (7a-c) contain the morpheme /-paʔa=/, ‘on’. 4 For
(7d), the morpheme containing the alternation is different, as there were no examples of pre-
3 In actuality, the discussion presented here of Comanche consonant mutations is necessarily oversimplified due to the simple fact that a more detailed discussion of these phenomena is outside the scope of this dissertation. The patterns, as they actually occur, are more nuanced than this and may depend on stress placement and other factors. The Comanche examples given here are simply meant to show a very basic picture of how consonant mutation processes occur in a Numic language. 4 Note that this morpheme itself is marked for a final feature: if another morpheme were added after /-paʔa=/ in any of these examples, the initial consonant of the following morpheme would likely surface as the plain stop variant due to the presence of the / =/ feature, which as noted in (6), triggers the appearance of a plain stop. 69
aspirated /-paʔa=/ that were found. However, the phoneme which undergoes the change in this example is still /p/, and the pattern still holds. The /p/ in /-paʔa=/ or, in the case of (7d), /-pitɨ/ undergoes mutation based on the final feature of the preceding morpheme. In (7a), the relevant morpheme is unmarked, which causes the following consonant to appear as a voiced spirant [v].
In (7b), the relevant morpheme is marked with the /=/ feature, causing the following consonant to surface as a plain stop [p]. In (7c), the noun stem is marked with the /h/ feature, causing the adjacent consonant to surface as a voiceless fricative [ɸ]. 5 Finally, in (7d), the initial morpheme in this example is marked with the /H/ final feature, which is associated with the pre-aspiration process. As such, the following consonant surfaces as a pre-aspirated [ hp].
In this analysis, if a Comanche speaker knows which final feature is present on a morpheme (or which ‘type’ it belongs to), then the initial consonant of the following morpheme can be easily predicted. This system, in Comanche and other Numic languages, is productive and regular. In Kawaiisu, however, the surface forms of word-medial, morpheme-initial consonants, cannot be predicted with the same regularity.
First, I consider the verb given by Zigmond et al. (1990) as /karü-/ ‘to sit’. When this morpheme occurs word-initially, it surfaces as [karü-], with the /k/ unchanged. In some instances, however, it is lenited to the spirant [ɣ], as in the form [ʔaːˈɣarü]
5 Charney (1993) notes that this may also trigger voicelessness on the vowel. This is shown in the (7c) form. She also notes that Comanche does have a phoneme /h/, but it her examples, it is clear whether she intends to use /h/ as a phoneme or as a final feature, because the phoneme /h/ only appears syllable-initially, while the final feature /h/ only appears syllable-finally. 70
[ʔaɣa-] ‘red’ must allow following consonants to surface unchanged, based on the behavior of the initial /k/ in the /karü-/ morpheme in the forms [ʔaːˈɣarü]
In fact, further examination seems to confirm this, at least at first glance. Zigmond et al.
(1990) note that, for example, the form [ʔaɣatübipaˈʔadü]
Zigmond et al. (1990) introduce the morpheme /kwiži/
[ʔaɣa-] morpheme, however, the lenited form surfaces: [ʔaɣaˈɣwiʃa]6
Also, reconsidering the form given above, [ʔaɣatübipaˈʔadü]
Canyon,’ note the unlenited [p] surfacing after the /tübi/ morpheme. One might take this as evidence that the /tübi/ morpheme allows a plain stop to surface in the following morpheme.
This, however, is not necessarily always the case. Consider the form /ˈkaːzi/
Without the influence of a preceding morpheme, this form surfaces as [kaːzi]. However, when preceded by the [tübi] morpheme, the form that appears is [tübiˈɣaːzi]
6 The unchanged form of this root, [kwiži], is different from the [ɣwiša] form in more than just the initial consonant. The second syllables in these forms differ from each other. However, Zigmond et al. (1990) still provide this as an example of the spirantization process without discussing why these forms have these additional differences. 71
(8) Unpredictability of Kawaiisu consonant mutations
Preceding Following Surface Orthographic Form Effect
(a) /ʔaː-/ /karü/ [ʔaːˈɣarü]
(b) /ʔaɣa-/ /karü/ [ʔaɣakaˈrüdü]
(c) /ʔaɣa-/ /tübi/ [aɣatübipaˈʔadü]
(d) /ʔaɣa-/ /kwiži/ [ʔaɣaˈɣwiʃa]
(e) /tübi/ /-paʔa/ [aɣatübipaˈʔadü]
(f) /tübi/ /kaːzi/ [tübiˈɣaːzi]
In (8b-d), the effect of the /ʔaɣa-/ morpheme is unclear. Sometimes it triggers spirantization, as in (8d), and other times it does not trigger a change at all. In (8e-f), the /tübi/ morpheme causes similarly unpredictable behavior. Sometimes, as in (8e), this morpheme does not result in a change, while in other instances, such as (8f), the spirantized form surfaces.
Additional unpredictable examples are given in Zigmond et al. (1990). With this information, it is not useful to attempt to categorize Kawaiisu morphemes as belonging to a certain ‘type’, as
Sapir (1931) does for Southern Paiute, or having certain final features which cause predictable effects, the way Charney (1993) analyzes Comanche.
While Zigmond et al. (1990) are clear about the unpredictability of medial consonants in
Kawaiisu compared to other Numic languages, they do hint that some patterns may exist, providing a chart showing the variant allophones of these consonants, shown below in (9).
72
(9) Kawaiisu Consonant Forms (Zigmond et al 1990:145)
Initial Consonant and Altered Medial Form
Unaltered Medial Form
“Spirantized” “Nasalized”
p v b (mb)
t r d (nd)
k g
kw gw
c z
č ž
The actual chart given by Zigmond et al. (1990) contains all of the Kawaiisu consonants.
Here, I only list the ones for which alternating forms are given. The consonant symbols used here are the same used in the consonant inventory proposed by Zigmond et al. (1990), shown previously in Table 1: the sound symbolized by /g/ is actually a voiced velar fricative, the sound symbolized by /č/ is a palatalized affricate, and the /ž/ represents the corresponding voiced fricative.
This chart shows that /p/ and /t/ actually have three variants: /p/ can occur as the plain
(unaltered) [p], spirantized [v], or as [b]. Zigmond et al. (1990) list [b] under the nasalized column, as they note that some speakers may produce this sound as [mb], with prenasalization.
The phoneme /t/ may appear as unaltered [t], spirantized [r] (varying freely between a flap and a trill), and nasalized [d] or [nd]. The remaining consonants in the chart have only two variants: the 73
unaltered forms [k], [kw], [c], etc., and their “altered” counterparts [g], [gw] (actually fricatives),
[z], etc .
While this chart is informative in that it shows the ways in which medial consonants in
Kawaiisu can surface, Zigmond et al. (1990) do not elaborate precisely on the conditions which govern the appearance of the various surface forms, writing the following:
“A full description of the lexically governed processes that determine the selection of
particular combining variants in synchronic Kawaiisu is beyond the scope of this work.
Often, grammatical morphemes which look like conditioned variants in fact have
different meanings. Even lexical morphemes do not always have a consistent effect on
following consonants. We have not attempted to mark each entry in the dictionary for its
“final feature”: we believe this concept is not uniformly valid for Kawaiisu, and in many
cases we simply do not have the data”
(Zigmond et al. 1990:145-146).
1.4.3. Morphology
Kawaiisu is a highly agglutinative language, with a strong tendency towards suffixes over prefixes. Words tend to be composed of several morphemes, with each morpheme encoding a specific piece of grammatical information. In an agglutinative language like Kawaiisu, each morpheme has one specific role. Consider the following simple Kawaiisu sentence given by
Zigmond et al. (1990:26):
74
(9) Kawaiisu agglutination
tami nana-paka-vaː-dü-mü
we.two RCPR -kill-IRR -NMR -PL
‘We two will fight.’
The form /nanapakavaːˈdümü/ is highly complex in terms of its morphology, composed of five discrete morphemes. In an agglutinative language like Kawaiisu, each morpheme contains only one piece of grammatical information. The morpheme /nana-/ is a reciprocal marker, indicating that the actors of the verb action are doing something to each other. The root /paka/ is glossed as ‘kill’. The suffix morpheme /-vaː/ indicates the irrealis mood, noting that the action described by the verb has not yet occurred. The suffix /-dü/ is called a nominalizer by Zigmond et al. (1990), but they also describe it as a generic marker that appears on verbs in main clauses.
Finally, the /-mü/ suffix indicates that the verb has a plural subject. In a fusional language, all this information could conceivably be packed into a single morpheme, but in an agglutinative language like Kawaiisu, each morpheme has one specific grammatical function, as shown by the example above in (9).
Nouns in Kawaiisu frequently occur with an absolutive suffix. Zigmond et al. (1990:39) list these as /-ci/, /-či/, /-zi/, /-ži/, /-bi/, /-pi/, /-vi/, /-bü/, /-pü/, /-vü/, /-büzi/, /-püzi/, and /-vüzi/.
Oberly (2008) notes, however, that the way many Uto-Aztecanists use the term absolutive has nothing to do with case marking in an ergative-absolutive system. Rather, it seems likely that the so-called absolutive suffixes behave like a noun classification system: each noun is lexically specified for a certain absolutive suffix. That is, speakers must know whether a given noun 75
belongs to the class which takes the /-bü/ suffix, or to the class which takes the /-zi/ suffix, etc.
This is similar to gender in other languages, or noun classes in Bantu languages, where speakers must simply know which category a given noun belongs to. Thus, it seems reasonable to draw similarities between the Uto-Aztecanist usage of the term ‘absolutive’ and gender/classification systems.
While the absolutive suffixes do often occur with nouns, these suffixes are sometimes dropped in some constructions, especially those involving plural suffixes, possession, and postpositions. Some examples are shown below in (10), taken from Zigmond et al. (1990).
(10) Nouns With and Without the Absolutive Suffix
(a) /taʔni-püzi/
man-ABS man-PL
(c) /nawa-bü/
tracks-ABS tracks=his
(e) /tüva-pü/
pinyon-ABS pinyon-like
The examples in the left column in (10) show the nouns with the absolutive suffixes, while the right column shows nouns with other suffixes. In (10a) and (10b), the absolutive suffix disappears when the noun is made plural. In (10c) and (10d), the contrast is between a noun with an absolutive suffix versus with a possessive clitic. (10e) and (10f) show that the absolutive suffix can be dropped when the noun is followed by a postpositional suffix.
The presence or absence of the absolutive suffix can have an effect on stress. Recall that
Kawaiisu stress falls on the penultimate mora. In (10a), this places stress on the /ü/ of the suffix 76
in /taʔni-püzi/. However, because the plural morpheme in (10b) is only one syllable, the final syllable of the root is stressed, the /i/ in /taʔni-mü/. A similar situation occurs in (10e) and (10f).
In (10e), stress falls on the final /a/ in the root /tüva/. However, in (10f), the penultimate mora of the word, and therefore stress, falls on the first /e/ the suffix /-reʔe/. Such pairs may facilitate the comparison of stressed and unstressed vowels.
1.4.4 Basic Word Order
As the main focus of this dissertation is on the phonetics of Kawaiisu, a fully detailed description of Kawaiisu syntax falls somewhat outside its scope, and can be found elsewhere, namely in Zigmond et al. (1990), which contains a thorough discussion of the syntactic structures of the language. Only a few basic details of Kawaiisu word order are provided here.
Typologically, Kawaiisu has several features commonly found in SOV languages. In syntactic theory, SOV languages are considered to be head-final, with right-branching syntactic structures. That is, the word that acts as the head of a syntactic constituent is often the last word in that constituent. Languages of this type tend to be strongly suffixing, and use postpositions
(rather than prepositions, as in English). Kawaiisu has both of these features: as discussed above, Kawaiisu tends to use suffixes, as opposed to prefixes, and postpositions come after their objects. Thus, Kawaiisu is similar to a typical SOV language in this regard. Zigmond et al.
(1990:14) provides an example of a simple Kawaiisu sentence showing SOV word order, shown below in (11).
77
(11) SOV Word Order in Kawaiisu
/taʔni-püzi momoʔo-a pükeː-rü=ina/
man-ABS woman-ACC see-NMR =her
‘The man saw the woman.’
In this example, the subject, /taʔni-püzi/ ‘man’ appears first, followed by the object,
/momoʔo-a/ ‘woman’, which is marked for accusative case. The accusative suffix /-a/ denotes this word as the object of the sentence. The verb comes last, /pükee-rü=ina/ ‘see’, which is marked with the verbal suffix /-rü/ 7, followed by the enclitic /=ina/, which denotes that the object of the verb is animate.
This word order is common in Kawaiisu. However, as the object of the Kawaiisu sentence is overtly marked for accusative case, speakers do not need to rely on word order to distinguish between subjects and objects: objects are determined by the presence of an accusative suffix. Since speakers can use this information to distinguish between subjects and objects, the sentence in (11) can be scrambled such that all word orders are possible: SOV, SVO, OVS, OSV,
VSO, or OVS.
In the Sheldon Klein recordings (Klein 1958, 1981-1983), used in the phonetic analyses presented in this dissertation, speakers often produce SOV sentences. However, when the subject of the sentence is a pronoun, it is often the case that the pronoun comes last, resulting in sentences like the one below, provided by Zigmond et al. (1990:15).
7 The exact function of this suffix is too complicated to go into detail here. Zigmond et al. (1990) label it as a nominalizing suffix, despite the fact that it occurs frequently on forms used as verbs. For a more complete discussion of this suffix, I direct the reader to chapter 4 of Zigmond et al. (1990). 78
(12) VOS Word Order in Kawaiisu
/pükeː-ka-dü=ina taʔni-püzi-a nüʔü/
see-R-NMR =him man-ABS -ACC I
‘I saw the man.’
In this example, the verb comes first. Some of the verbal morphology here is the same as the verb in (11), but see Zigmond et al. (1990) for a more detailed discussion of the morphology of Kawaiisu. The second form, /taʔni-püzi-a/ ‘man’, is similar to /taʔni-püzi/ in (11), but here it receives the accusative suffix /-a/, marking it as the object of the sentence. The pronoun /nüʔü/
‘I’ occurs in final position.
1.5. Data and Methods The audio data that forms the basis of the acoustic analyses presented in this dissertation come from The Sheldon Klein Collection of Kawaiisu Sound Recordings, hosted online at the
California Language Archive (Klein 1958, 1981-1983). Criteria for the specific measurements of each type of sound will be given in each chapter, but general methods are discussed here. In this section, I provide a general overview of what Kawaiisu data are hosted in the California
Language Archive, discussing their practical applicability to phonetic analysis, as well as some limitations.
Sheldon Klein’s (1958, 1981-1983) Kawaiisu audio data contain recordings from 1958, and the early 1980s. In each of these eras, he worked with different speakers. Speakers from
1958 do not appear in the recordings from the 1980s, and vice-versa. The recordings themselves 79
are taken from over 100 different tapes, split into approximately 135 digital audio files. The files, which can be accessed directly from the California Language Archive website, are downloadable in .wav format. Information about the specific audio recording equipment used by Klein during his data collection is apparently unavailable, although it is likely that the equipment used in 1958 is not the same as the equipment used in the 1980s. Information regarding the process by which the tapes were digitized is also not obviously available, save for a note that the digitization itself was sponsored by the National Endowment for the Humanities. In spite of the apparent lack of information regarding recording equipment and digitization, it is fortunate that the files themselves are hosted in .wav format, as this format typically contains uncompressed audio.
Other file formats, such as .mp3, may have smaller file sizes, but are generally highly compressed resulting in a potential loss in audio quality.
Several different types of data are represented in this archive. Most of the files included in the archive were recorded in person, but a few tapes include interviews and conversations recorded over the telephone. Klein’s Kawaiisu files contain discussions of Kawaiisu cultural phenomena, often in English, but some stories are told in Kawaiisu as well. Some of the tapes include monologues in Kawaiisu and conversations between speakers in the language, and at least one file is reported to contain a traditional recipe. In this dissertation, I use the speech analysis software package PRAAT (Boersma and Weenink 2017) to examine those files which contain elicitation sessions of Kawaiisu lexical items and grammar. I do not analyze the conversations, story-telling, or the recipe because of the impossibility of determining what each word is without the assistance of a native speaker collaborator, in such connected speech files without transcription. In these tapes of elicitation sessions, the interviewer (usually Sheldon 80
Klein) asks the speaker (or speakers) the Kawaiisu word for a familiar concept, and the Kawaiisu speaker responds with the appropriate Kawaiisu word. For instance, the interviewer may ask for the Kawaiisu word for “moon,” and the Kawaiisu speaker responds with the Kawaiisu word for this concept, “müazi.”
After the speaker pronounces each word, I use the Zigmond et al. (1990) dictionary to check each word, making sure the word and all its morphemes can be found in the dictionary before labeling the word in Praat. This step is necessary for a few reasons. First, English contains many homophones, such as “son” and “sun.” For example, Klein may be attempting to elicit the word for “son,” while the speaker responds with the Kawaiisu word “tavi,” referring to the sun, rather than one’s male child. Other times, Klein asks for a word, and the speaker responds with a different word with a somewhat similar meaning. For instance, in one tape, Klein asks for directional terms like “east,” “north,” etc., and the speaker gets the Kawaiisu directional terms mixed up. By checking the speakers’ words using the Zigmond et al. (1990) dictionary, some of the confusion is avoided, allowing for more of the data to be examined.
In many of the tapes, Sheldon Klein assumes the role of interviewer-linguist, asking questions of the Kawaiisu speakers and leading the discussions. However, a few tapes involve elicitation of words and cultural discussions relating to childbirth and female sexuality. In these instances, the role of interviewer is assumed by Sheldon Klein’s wife, Carol Klein, who discusses these potentially sensitive topics with female Kawaiisu speakers.
In total, Sheldon Klein’s (1958, 1981-1983) Kawaiisu audio archive contains recordings of 10 Kawaiisu speakers. Some of these speakers, however, did not participate in the elicitation sessions, and therefore their data is not analyzed in this dissertation. Of the original 10 speakers 81
who were recorded by Klein, this dissertation considers data from 4 speakers: Fred Collins, Anne
Peltier, Lida Girado, and Rose Barneche (henceforth FC, AP, LG, and RB, respectively) 8. These speakers were listed by name on the California Language Archive website, so they are credited for their work by name here as well. The male speaker (FC) was recorded in 1958, and the three female speakers (AP, LG, and RB) were recorded in the 1980s. Klein does not provide the ages of the Kawaiisu speakers he worked with, but he does include a few photographs. Examining the photographs, it is apparent that the speakers (at least the four considered in this dissertation) were all middle-aged or older at the time of elicitation.
While the elicitation session audio from these four speakers does provide much useful data for phonetic analysis, there are some drawbacks and limitations. First, as previously mentioned, information is not available regarding the digitization process or Klein’s recording equipment. Differences in recording equipment can explain, among other things, differences in audio quality, which can possibly have an effect on the results of the analyses presented in this dissertation. Second, differences in audio equipment are not the only reason some files may differ in audio quality. Some tapes seem to have been recorded outdoors, with wind blowing, birds chirping, and cars driving by. Even those tapes which seem to have been recorded indoors sometimes vary in audio quality. On some tapes, speech is clear, while on other tapes, speakers respond to Klein’s queries while chewing gum, eating, or holding their hands over their mouths, which does not lead to easily analyzable recorded speech (Klein, himself, sometimes asks speakers to quit chewing gum, or move their hands away from their faces). Some tapes include
8 A fifth speaker, Rose Collins, does participate in some of the elicitation sessions from 1958. However, her recordings are not analyzed in this dissertation for two reasons. First, she does not actually provide much data, giving only a dozen or so words. Second, the audio quality on the tapes which include her is so poor that the small amount of data she does provide cannot be analyzed. 82
background noise from the television, radio, and other home appliances. Some tapes, recorded in speakers’ homes, include noise from children running through the house and playing. Some tapes which would otherwise be fine contain a considerable amount of tape hiss.
Aside from this, each speaker is given different lists of words and phrases for elicitation.
This makes the data somewhat messy and difficult to carefully analyze, for a few reasons. First, because no two speakers have the same word lists, it is not always possible to compare different speakers’ pronunciations of the same words. Second, some types of sounds may be relatively overrepresented in some speakers’ data, but underrepresented or even absent from the data of others. Additionally, in an ideal phonetic study, it is advisable to consider the phonological environment of the sounds being analyzed. For example, when studying the quality of vowels, one may wish to analyze vowels only in a certain consonantal environment, as various consonantal places of articulation may affect a vowel’s acoustic signature in different ways
(Ladefoged 2003). In a study such as this one, based on limited archival data from a severely endangered language, the phonological environment of sounds under analysis cannot be controlled.
This discussion of the data used in this dissertation should not be taken as an apology to the reader, but rather as an explanation for the benefit of those who may not be familiar with research on severely endangered languages. When studying a widely spoken language like
English, researchers are more likely to have access to large numbers of highly fluent speakers, and can manipulate word lists or experimental stimuli with the aim of studying specific linguistic phenomena. When dealing with severely endangered languages, on the other hand, researchers may only have access to a few speakers who may only remember bits and pieces of the language, 83
making it much more difficult to design tightly controlled experiments and word lists. The nature of the data collected in a study on a language like Kawaiisu is necessarily different from the data that can be collected in a study on languages with huge populations of native speakers. As such, different approaches and methods must be used when dealing with endangered language data, which may seem unusual to those who do not regularly deal with such languages. When there are only a handful of L1 speakers, any data is good data.
1.6. Organization of the Dissertation This dissertation contains six chapters. Chapter One has provided an introduction to the
Kawaiisu language, giving an overview of prior research on the language, some of its grammatical structure, its genetic affiliation, and state of endangerment, as well as discussing the data used in the remainder of the dissertation. Chapter Two is an examination of the stressed vowels of Kawaiisu. Chapter Three discusses the acoustics of unstressed Kawaiisu vowels. In
Chapter Four, the data from the previous two chapters is used to discuss the acoustic correlates of stress in Kawaiisu, comparing the acoustics of Kawaiisu stress to that of Southern Ute. In
Chapter Five, I provide an acoustic analysis of Kawaiisu consonants. Final conclusions and discussions are contained in Chapter Six. 84
CHAPTER 2: STRESSED VOWELS
2.1 Introduction This chapter presents an in-depth examination of the Kawaiisu stressed vowel space. As noted in the previous chapter, stress in Kawaiisu tends to occur on the penultimate syllable. The acoustic cues for stressed vowels are often more robust than those for unstressed vowels, and as such, provide an excellent starting point for a detailed examination of the Kawaiisu vowel space in general. I examine the unstressed vowel space in Chapter 3, and in Chapter 4, the stressed vowels and unstressed vowels are compared in a discussion of the acoustic correlates of
Kawaiisu stress.
Kawaiisu has six phonemic vowel qualities. Additionally, there is a quantity distinction: each vowel can be phonemically long or short. An examination of the lexicon included in
Zigmond et al. (1990) reveals some minimal pairs. One such example is /ˈkwija/
The six phonemic vowel qualities of Kawaiisu are described in varying levels of detail by
Klein (1959, 2002), Zigmond et al. (1990), and Thomas (2017). The differences between the analyses of the Kawaiisu vowel space presented by Klein (1959, 2002), and by Zigmond et al.
(1990) were discussed in Chapter 1. Essentially, these sources agree in their phonetic descriptions of five of the vowels: /a, e, i, o, u/. The disagreement lies with their ideas of the 85
sixth vowel. Zigmond et al. (1990) describe the sixth vowel as a high, back, unrounded /ɯ/, while Klein (2002) argues that the vowel is better described as a high central /ɨ/, while also noting that the vowel may be somewhat lower in unstressed environments. Thomas (2017) discusses the acoustic quality of this vowel, arguing that it should be analyzed phonetically as a central vowel whose height ranges between high and mid, proposing that the phonetic symbol
[ɘ], standardly used for an unrounded mid-high central vowel, is the most appropriate representation. In this dissertation, the symbol /ü/ is generally used, as this is the symbol used in the orthography adopted by the community. However, when quoting earlier work on Kawaiisu, I attempt to use the symbols used by the original authors.
While Thomas (2017) is in agreement with Klein (2002) on the phonetic centrality of this vowel, I present phonological arguments in favor of the Zigmond et al. (1990) analysis of this vowel as a high, back, unrounded vowel. Thus, Klein’s (1959, 2002) description may be more accurate from a phonetic point of view, but Zigmond et al. (1990) are correct from a phonological perspective.
Thomas (2017) examines vowel coalescence data provided by Zigmond et al. (1990) and overviewed briefly in Chapter 1 of this dissertation. Essentially, at a clitic boundary, when the stem ends in a vowel, and the following clitic begins with a vowel, vowel coalescence occurs, a process by which the two separate short vowels merge into a single long vowel, sharing features of both of the underlying short vowel phonemes. In his argument, Thomas (2017) invokes
Casali’s (1996) Feature Preservation Preference, which states that if either member of a vowel sequence undergoing coalescence is underlyingly non-high, the long vowel which surfaces will also be non-high. It is shown in Zigmond et al. (1990), and in Chapter 1 of this dissertation, that 86
when concatenation results in an underlying /üi/ sequence, the result is a long /iː/. In order for
Casali’s (1996) Feature Preservation Preference to hold, this means that the /ü/ vowel must be regarded as phonologically high, as the surface result of vowel coalescence in this case is a long high vowel. If /ü/ were not a phonologically high vowel, then /iː/, according to Casali (1996), could not surface in this position. In terms of Distinctive Feature Theory (Jakobson et al. 1952,
Chomsky and Halle 1968), this vowel can be analyzed as having the [+high] feature.
Further, in Chapter 1 of this dissertation it was noted that coronal obstruents become palatalized following a front vowel. There, the form [ʔüˈkütsi] <ükütsi> ‘little hat’, was provided, where the [ts] in the diminutive suffix /-tsi/contrasts with a palatalized [tʃ] in forms like [ʃivoroˈnitʃi]
(1990) argue that the palatalized allophone appears due to the influence of the preceding [i], a front vowel. If palatalization is triggered by a preceding front vowel, and [ʔüˈkütsi] <ükütsi> does not surface with a palatalized allophone, then the /ü/ cannot phonologically be a front vowel. In terms of distinctive features, this shows that the vowel has the feature [+back].
The Kawaiisu processes of palatalization and vowel coalescence show that the /ü/ vowel can be regarded as a high back vowel (with the features [+high, +back]), however, this does not account for the third feature commonly ascribed to vowels: the rounding feature. Thomas (2017) argues that if the sixth Kawaiisu vowel is [+high] and [+back], then it is necessarily [-round].
The reason for this argument is that Kawaiisu already has a vowel which unambiguously has the features [+high, +back, +round]. That vowel is /u/. In order for these vowels to be phonologically distinct, they cannot have the exact same set of feature specifications, and Klein (2002) seems to indicate that this vowel should not be regarded as a round vowel in the first place. Based on the 87
arguments presented by Thomas (2017), I regard this vowel as being the high, back, unrounded vowel phoneme /ü/. The Kawaiisu vowel phonemes are listed below in Table.
TABLE 1: Kawaiisu Phonological Vowels Front Vowels Back Vowels [-back] [+back] Unround Vowels Round Vowels [-round] [+round] High vowels [+high, -low] i ü u Mid vowels [-high, -low] e o Low vowels [-high, +low] a
As noted, vowel length is phonemic in Kawaiisu. In some languages said to have phonemic vowel length, vowels which differ in length may also differ in quality as well. As such, I examine long and short Kawaiisu vowel pairs in this chapter to test for differences in vowel quality.
In Comanche, a language closely related to Kawaiisu which is also reported to have phonemic vowel length, Herrick (2011) notes that some speakers may have slight qualitative differences between some of their long and short vowel pairs, although, for the most part,
Comanche vowel length is solely a matter of duration, rather than quality. In her study of
Southern Ute, Oberly (2008) does not note differences in quality between long and short vowel pairs, but does give some examples showing how long and short vowels differ in duration. In this chapter, I compare the formant structures of long and short Kawaiisu vowel pairs to test for differences in vowel quality.
88
2.2 Methods In the remainder of this chapter, I examine the duration and formant structure of stressed vowels in Kawaiisu to determine what differences exist between long and short stressed vowels in this language. If a vowel is listed in the Zigmond et al. (1990) dictionary as a long vowel, I consider it to be a long vowel for the purposes of this study. Vowels listed as short in Zigmond et al. (1990) are considered short here as well. Long and short vowels are tested for differences in duration, as well as for differences in vowel quality. The data for this chapter come from Sheldon
Klein’s Kawaiisu field recordings (Klein 1958, 1981-1983), discussed in section 1.5 of this dissertation. All tokens examined come either directly from the wordlist elicitation sessions, or the sentence elicitation sessions. In cases where a token comes from the recordings which include elicitation of full sentences, the words examined were uttered in isolation, rather than being part of one of the elicited sentences. Usually this situation arises when Klein asks the speaker for a Kawaiisu sentence, which is then produced, and Klein has clarificatory questions about one of the words in the sentence. None of the tokens chosen for examination in this chapter come from connected speech. Although much can be learned from studying full sentences or phrases or other types of connected speech, I choose to omit these types of speech from the present study because, cross-linguistically, vowels tend to be reduced in conversational speech, rapid speech, or longer utterances (Koopmans-van Beinum 1980), and because it is best to examine a phenomenon that has not been studied before in careful speech first. Examining only vowels from words produced in isolation allows us to produce a clearer picture of the Kawaiisu vowel space. A description of the Kawaiisu vowels which mixes different speech types and speech rates would not yield meaningful results, and would not produce a very informative picture of what the vowels of the language are like. Additionally, examining connected, rapid, or 89
otherwise conversational speech would be prohibitively difficult without the collaboration of a native Kawaiisu speaker.
All vowel tokens examined in this chapter, both short and long vowels, are stressed vowels from the penultimate syllable of the word. This is the normal position for Kawaiisu lexical stress. Words which end in a long vowel have stress on the final syllable, however, none of the tokens examined in this chapter have word-final stress. 9 There are a few reasons for omitting such words from consideration here. First, such words are comparatively rare in
Kawaiisu. Only a few examples exist in the Zigmond et al. (1990) dictionary. Second, due to the nature of the data, many important phonological factors could not be controlled in a consistent way. For example, surrounding consonantal environment could not be tightly controlled, and as such, the effect of surrounding consonantal environment on factors such as vowel length could not be measured. However, limiting the vowels under examination to penultimate syllables only is one thing that can be controlled, resulting in a set of data that is more uniform and clear than it otherwise could have been. Finally, as mentioned, this chapter examines the duration of long versus short vowels. Cross-linguistically, there is a tendency for word-final vowels to be held for a greater duration than vowels elsewhere in a word (Delattre 1966, Oller 1973, Oller & Smith
1977). Therefore, even if word-final stressed vowel data were easily available in Kawaiisu, including such tokens would likely have the effect of making the data unnecessarily difficult to interpret due to the effect of word position.
9 Zigmond et al. (1990) analyze Kawaiisu stress as falling on the penultimate mora, and in words with final long vowels, the penultimate mora is in the final syllable, since long vowels are considered to be bimoraic. A discussion of this analysis is included in Chapter 1 (Section 1.4.2.3) of this dissertation. 90
Because of the way each speaker’s wordlists vary, the vowel tokens here represent a wide range of surrounding phonological environments. The only possible environment not represented in the data I examine is the pre-nasal environment. That is, vowels which are followed immediately by /m/ or /n/ are not examined here, due to the effect that pre-nasal environments have on the preceding vowel (discussed, for example, in Johnson 2012). In order to provide a general description of the Kawaiisu vowel space, it is necessary to omit this environment from consideration. However, all of these omissions (pre-nasal environments, word-final long vowels, conversational speech rate, and longer utterances) may prove to be interesting topics for future research.
Vowels in post-nasal environments, on the other hand, were considered. Vowel tokens selected for analysis in this chapter include those which follow nasals, obstruents, and glides.
They can be followed by either obstruents or glides. All places of articulation found in Kawaiisu are included in the pre- and post-vocalic environments examined in this study: labials, coronals, velars, and labiovelars. All of these places of articulation occur with similar frequency, so any effect place of articulation may have on vowel formant structure should be averaged out.
For this portion of the study, each word was labeled individually using Praat software
(Boersma and Weenink 2017). After labeling each word and determining the locations of the starting and ending points for each stressed vowel token, a Praat script was used to extract the duration of each vowel, as well as the fundamental frequency (pitch), and F1-F3 values at 20%,
50%, and 80% of the duration. The way starting and ending points for vowels were defined differed based on surrounding phonetic context. In cases where the vowel was bordered by obstruents, the starting point of the vowel was taken to be the onset of F2, and the ending point 91
of the vowel was taken to be the offset of F2. There were some instances, however, in which F2 was weakly present during an adjacent consonant. In these instances where F2 weakly appears on a preceding consonant, the beginning point of the vowel is taken to be the point where F2 most suddenly gains amplitude, and in cases where F2 weakly appears on a following consonant, the endpoint of the vowel is taken be the point where F2 most suddenly loses amplitude. An example is shown below in Figure 1.
FIGURE 1: Waveform and spectrogram of /keːvi/
Although vowels in pre-nasal environments were not measured, post-nasal vowels were.
In post-nasal environments, release of a nasal consonant into a following oral vowel is signaled 92
by an abrupt shift in frequency distribution (Hayward 2000). These sudden shifts of energy are quite noticeable on a spectrogram, and as such, I count the abrupt shift from nasal formants into following oral vowel formants as the onset of a post-nasal vowel. In other words, the abrupt shift in F2 frequency is taken to be the starting point of post-nasal vowels.
Other issues must be considered when defining vowel boundaries when the adjacent environment contains a glide such as /w/ or /j/. Glide consonants, often vowel-like, have formants which are visible in the spectrogram, unlike obstruents. This means that onset or offset of F2 cannot be used to determine the boundaries of the vowel. In fact, Ladefoged (2003) advises against attempting to determine boundaries between glides and vowels altogether, precisely because no clear phonetic boundary exists as the glide transitions into the phonological vowel. In these situations, boundaries were placed at the specific point in the waveform where a sudden change in amplitude is apparent. If this visual cue was not present, then the vowel was not selected for analysis.
For example, as /w/ is similar articulatorily and acoustically to /u/, and /j/ is similar to /i/, the formant pattern for /w/ is associated with a low F2, and /j/ with a high F2. As speakers make the articulatory transition between the glide and the vowel nucleus, the formants move from the frequencies associated with the glide to the frequencies associated with the vowel. An example is shown below in Figure 2. 93
FIGURE 2: Waveform and spectrogram of /naˈvojo/
Here, the stressed /o/ in the penultimate syllable in this example is bounded on the left by
/v/, an obstruent, and on the right by /j/, a glide. The left edge of the boundary in the ‘stressV’ tier is aligned with the onset of F2, using the method described above for obstruent-containing environments. Following the vowel, however, is /j/, a glide. Note how the F2 for /o/ begins fairly low, as expected for a back vowel. Throughout the production of the /o/, F2 rises during the transition into /j/, the palatal glide, as /j/ is similar to /i/, a front vowel with a high F2. In the textgrid, the right edge of the boundary appears slightly before F2 reaches the maximum frequency for the /j/. The waveform also indicates that the amplitude for the stressed /o/ is decreasing at this point, as the higher intensity for the stressed vowel gives way to the consonantal glide /j/.
94
2.3 Results 2.3.1 Duration Analyses
Using the methodology outlined above, a total of 1,252 stressed vowel tokens were analyzed. A chart showing the number of stressed vowel tokens from each speaker is shown below in Table 2.
TABLE 2: Number of stressed vowel tokens available for analysis for each speaker and vowel phoneme (long and short vowels listed separately). This represents the number of each stressed vowel in the phonological environments described for inclusion. Speaker Vowel phoneme
a aː e eː i iː o oː u uː ü üː
AP 57 29 2 53 113 51 16 2 13 4 42 8
FC 77 16 1 22 61 14 16 2 22 5 27 5
LG 87 18 4 21 124 21 24 4 23 2 68 8
RB 29 23 0 23 40 21 10 1 4 1 34 4
Total 250 86 7 119 338 107 66 9 62 12 171 25
Below, Figure 3 shows the difference in duration between phonemically long and short vowel pairs.
95
FIGURE 3: Comparison of long and short vowel durations in stressed position
Each bar in the figure above represents the mean duration for each vowel across all speakers. In every case, the phonemically long vowels (represented by green bars) are longer in duration than the phonemically short vowels (represented by blue bars).
For each speaker, the average duration of the phonemically long member of a vowel pair is greater than its short counterpart. Short vowels in stressed position, on average, are around 60-
70ms, while the average long vowel duration in a similar position ranges from 110ms (in the 96
case of /iː/) to 213ms (in the case of /oː/). Some of the vowels of speaker LG (both long and short) tend to be somewhat greater in duration than the comparable vowels of other speakers.
Speech rate may play a role in this, but the main reason for this difference is that many of the tokens represented here come from examples where Sheldon Klein asks speakers to repeat a word that they have already said. In cases where LG was asked to repeat a word that was already said, she often drew the word out, holding the vowels for a greater period of time. It is possibly the case that she did this in order to make sure she was enunciating clearly so that Klein would be sure to record her pronunciation correctly in his notes, although she does not state specifically that this was her goal.
While stressed long vowels in Kawaiisu have a greater duration than stressed short vowels, the difference in duration is not the same for each vowel pair. From Figure 3, the difference between /i/ and /iː/ appears to be quite small, while the difference between /o/ and /oː/ is much greater. To more closely examine the relationship between duration and phonemic length of stressed Kawaiisu vowels, some statistical analyses were performed.
Because there were no tokens of short /e/ produced by speaker RB that could be included in the analyses, separate ANOVAs were conducted. In the first test, data were analyzed using a two-factor within subjects ANOVA with duration in milliseconds as the dependent variable. The two factors were phonemic length (two levels: long and short) and vowel type (five levels, one for each vowel type excluding /e/, as no tokens of short /e/ from speaker RB could be included in the analysis). In this test, there was a significant main effect of phonemic length ( F(1, 3) =
99.082, p = 0.002)), indicating that the difference in duration between long and short vowels is statistically significant. There was also a significant main effect of vowel type ( F(4, 12) = 97
15.639, p < 0.001), indicating that some vowels simply have a greater duration than others regardless of phonemic length (this is consistent with the descriptive results above which shows a much greater duration for /o/). The interaction of phonemic length and vowel type was also significant ( F (4,12) = 1682.709, p = 0.006), indicating that the difference in phonemic length between long and short stressed vowel pairs is realized differently depending on vowel type.
To examine this interaction, several pairwise comparisons were conducted comparing vowel types of like phonemic length. That is, comparisons were made between /a/ and /i/, /a/ and
/o/, /a/ and /u/, etc. A significant difference in duration was found between /a/ and /ü/ ( t(3) =
3.186, p = 0.05), fitting with the fact that /ü/ has a relatively short duration compared to the other phonemically short vowels.
Similar pairwise comparisons were conducted examining the long vowels. Significant differences were found between /aː/ and /iː/ ( t(3) = 5.066, p = 0.015), /aː/ and /oː/ ( t(3) = -4.216, p = 0.024), /iː/ and /oː/ ( t(3) = -8.950, p = 0.03), /oː/ and /uː/ ( t(3) = 3.819, p = 0.032), and /oː/ and /üː/ ( t(3) = 5.707, p = 0.011). Significance was approached for the /iː/, /üː/ pair, but not attained ( p = 0.051).
Short /e/ and long /eː/ were tested separately due to a lack of short /e/ tokens from RB.
This was a within-subjects ANOVA with two factors: phonemic length (2 levels: long and short) and vowel-type (one level: /e/). A significant main effect of phonemic length was found ( F(1, 2)
= 196.933, p = 0.005). Additional pairwise comparisons were conducted, comparing short /e/ to the other short vowels, and long /eː/ to the other long vowels. RB’s data were excluded from these comparisons. No significant differences were found in these comparisons.
98
2.3.2 Segmental Results
In addition to duration, the script used to measure vowels extracted pitch information and values for the first three formants at 20%, 50%, and 80% of the duration of each vowel. The results of these measurements are discussed in this section. Because of the necessarily small number of speakers, each was studied separately instead of attempting to normalize their vowel spaces. Below, Figure 4 shows an F1 x F2 vowel plot for the stressed vowels of FC.
FIGURE 4: F1 x F2 Vowel Plot from Speaker FC, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
99
Each vowel quality occupies an area in the vowel space which is reasonably distinct, although short /u/ and short /o/ both appear to occupy areas that are fairly close to each other.
Generally, the long vowels are shown to be more peripheral than their short counterparts: long
/iː/ is both higher and fronter than short /i/, long /üː/ is higher in the vowel space than short /ü/, long /aː/ is closer to the lower edge of the vowel space while short /a/ is more centralized along the F1 dimension, and long /uu/ appears higher than short /u/. The only phonemically long vowels which do not seem to occupy a more peripheral area than their short counterparts are /eː/ and /oː/. Short /e/ and long /eː/ show a high degree of overlap in the vowel space, while long /oː/ is slightly lower than short /o/, but are similar along the F2 dimension.
In addition to F1 x F2 vowel plots, like the one shown above, F1 x F3 vowel plots were also created for each speaker. 100
FIGURE 5: F1 x F3 Vowel Plot from Speaker FC, with F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
To investigate these data further, some statistical tests were conducted examining the effects of phonemic length, timepoint, and vowel type on F1, F2, and F3. First, by-items
ANOVAs were conducted on speaker FC’s vowel data to test whether long and short vowels differ from each other in quality. Each vowel pair was tested separately, so, for example, short /a/ was compared to long /aː/, short /i/ was compared to long /iː/, etc. The dependent variables in these tests were F1, F2, and F3, averaged over timepoint such that each token had a single value for each formant which was the mean of the formant measurements at 20%, 50%, and 80% of the 101
vowel’s duration. In each of the subsequent tests in this chapter, vowels with fewer than 5 tokens are not included in the statistical analyses. F-ratios and significance are shown below in Table 3.
TABLE 3: F-ratios and significance levels comparing long and short vowel quality, for each vowel, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 91 30.739*** 8.133** 3.368 /e/ Excluded from analysis due to paucity of data /i/ 1, 73 6.409* 3.376 10.504** /o/ Excluded from analysis due to paucity of data /u/ 1, 25 5.359* 1.161 1.012 /ü/ 1, 30 4.193* F<1 3.091 * p < .05 ** p < .01 *** p < .001
Above, Table 3 shows that short /a/ and long /aː/ are significantly different in terms of F1 and F2. Short /i/ and long /iː/ are significantly different in terms of F1 and F3. Short /u/ and long
/uu/ are significantly different in terms of F1, while short /ü/ and long /üː/ also differ significantly in terms of F1 only. Vowel types /e/ and /o/ were not included in this analysis because so few of these tokens were produced by this speaker. The results of these analyses are consistent with the F1 x F2 and F1 x F3 plots shown in Figures 4 and 5, respectively. The general trend is that long vowels are more peripheral in the vowel space than the short vowels, especially in the F1 and F2 dimensions.
In addition to testing long and short vowel pairs for differences in quality, tests were also conducted to determine how each vowel moves around the vowel space during its articulation.
Because measurements were taken for each vowel token at 20%, 50%, and 80% of the duration of each token, testing each vowel phoneme for differences in timepoint between F1, F2, and F3 can provide information regarding how diphthongal or monophthongal each phoneme is. That is, 102
if there are no significant differences between these three timepoints for these three formants, the vowel is taken to be monophthongal. Significant differences between timepoints are likely to indicate diphthongization, as vowel quality is shown to differ from one timepoint to the next. To investigate this, several repeated measures ANOVAs were conducted, with timepoint as a within-subjects factor (because each vowel has measurements at all three timepoints, each token is essentially counted as a subject with F1 measurements at 20%, 50%, and 80%, F2 measurements at 20%, 50%, and 80%, etc. ) F-ratios and significance are shown below in Table
4.
TABLE 4: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for FC. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 152 17.663*** 3.129* 2.994 /aː/ 2, 30 8.247** 9.212* 2.460 /e/ Excluded from analysis due to paucity of data /eː/ 2, 42 3.303* 4.624* F<1 /i/ 2, 120 11.559*** 10.221*** 12.981*** /iː/ 2, 26 10.376*** F<1 5.983** /o/ 2, 30 11.772*** 4.409** F<1 /oː/ Excluded from analysis due to paucity of data /u/ 2, 42 F<1 2.505 1.004 /uː/ 2, 8 F<1 F<1 2.207 /ü/ 2, 52 3.365* 4.997* 3.831* /üː/ 2, 8 F<1 3.653 1.030 * p < .05 ** p < .01 *** p < .001
Above, Table 4 shows that several vowels produced by speaker FC were articulated with at least some consistent movement. For example, timepoint has a significant effect on all three measured formants for short /i/, which is consistent with the information in Figure 4, which 103
shows short /i/ beginning (at 20%) with an F1 of approximately 425Hz and an F2 of approximately 1650Hz, and ending (at 80%) with an F1 of approximately 375Hz and an F2 of approximately 1750Hz. Short /i/ of speaker FC, therefore, moves higher and further forward in the vowel space throughout its articulation. Although Table 4 shows that many of speaker FC’s vowels are produced with at least some movement, the degree and direction of movement in the vowel differs from phoneme to phoneme. This can be seen in an examination of Figure 4. As discussed, short /i/ moves higher and further forward, while long /üː/ moves further forward in the vowel space throughout its articulation, but remains roughly the same in terms of height, and long /aː/ moves lower in the vowel space from 20% to 50%, while ending up higher and further back at 80%. There is no consistent pattern for direction of movement, but it is clear that timepoint does have an effect on the formants of many of FC’s vowel phonemes.
In the following pages, vowel plots and statistical tables like the ones seen for FC above are shown for the other Kawaiisu speakers whose data were analyzed in this study. For each of the following speakers, the same methods for data collection were used, and the same types of statistical analyses were used. Below, Figure 6 shows the F1 x F2 vowel plot for speaker AP. 104
FIGURE 6: F1 x F2 Vowel Plot from Speaker AP, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
The F1 x F2 vowel space of AP is similar to that of FC: each vowel seems to occupy a distinct area within the vowel space, with greater potential for overlap in the back vowels than the front. Further, it appears that the long and short vowel pairs seem to occupy slightly different areas within the vowel space; that is, there does not appear to be much overlap in acoustic quality 105
between the long vowels and their short counterparts. In every case, the long vowels are closer to the periphery than the short vowels. This is somewhat similar to FC’s long/short vowel pairs, but the difference is greater in the case of AP. Finally, note that AP’s long /oː/ seems to be somewhat unstable, showing an apparent glide to a higher position in the vowel space at 80% duration. In AP’s stressed vowel data, only two usable tokens of long /oː/ were produced (see
Table 2), making any conclusions about these vowels tentative. For each speaker, long /oː/ and short /e/ are the rarest vowels, so these vowels may seem somewhat aberrant in the vowel plots.
Below in Figure 7, a F1 x F3 vowel plot is presented for speaker AP. 106
FIGURE 7: F1 x F3 Vowel Plot from Speaker AP, featuring F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
When considering the F1 x F3 vowel plot of speaker AP, there appears to be two distinct regions in the F1 x F3 space: high vowels (especially long high vowels) lie in the upper left of the plot, while non-high vowels tend to fall mostly in the lower right of the plot.
These data were analyzed with the same tests used to analyze the vowel data of speaker
FC, detailed above. Below, Table 5 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair. 107
TABLE 5: F-ratios and significance levels comparing long and short vowel quality, for each vowel, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 84 18.496*** 1.735 F<1 /e/ Excluded from analysis due to paucity of data /i/ 1, 162 9.209** 11.261** 7.862** /o/ Excluded from analysis due to paucity of data /u/ Excluded from analysis due to paucity of data /ü/ 1, 48 2.510 2.989 4.555* * p < .05 ** p < .01 *** p < .001
The results of these analyses are largely consistent with the vowel plots shown in Figures
6 and 7 above. Long /aa/ is lower in the vowel space than short /a/, consistent with the significant difference in F1 for this vowel pair. Long /iː/ tends to be higher and further forward in the vowel space than short /i/, which is reflected in the significant differences found in F1 and F2 for this pair. This speaker did not produce many tokens of short /e/ or long /oː/, but a cursory examination of Figure 6 shows that long /eː/ and short /e/, as well as long /uː/ and short /u/, may be reasonably distinct, while long /oː/ and short /o/ seem reasonably close together. Generally
(with the possible exception of long /eː/, which was not included in this analysis due to paucity of data), the trend is that long vowels are closer to the edge of the vowel space than short vowels.
Below, Table 6 shows F-ratios and significance levels for effect of timepoint on formant structure, reflecting some movement throughout articulation. 108
TABLE 6: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker AP. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 112 F<1 F<1 2.150 /aː/ 2, 56 1.497 2.373 1.995 /e/ Excluded from analysis due to paucity of data /eː/ 2, 104 3.841* 3.091* 1.759 /i/ 2, 222 8.494*** 3.741* 1.206 /iː/ 2, 102 21.486*** 1.946 1.517 /o/ 2, 30 4.611* 4.180* 1.052 /oː/ Excluded from analysis due to paucity of data /u/ 2, 24 3.311 F<1 F<1 /uː/ Excluded from analysis due to paucity of data /ü/ 2, 82 3.216* 4.667* 1.710 /üː/ 2, 14 F<1 F<1 F<1 * p < .05 ** p < .01 *** p < .001
The findings shown above in Table 6 indicate that some of speaker AP’s vowels show at least some movement in the vowel space throughout their articulation. Long /eː/ starts in a roughly central position before moving forward and slightly higher in the vowel space. Short /i/ shows a similar pattern of movement, going higher and further forward throughout the articulation. Long /iː/ is slightly different, moving higher in the vowel space, but somewhat further back, while short /ü/ becomes lower along the F1 dimension and more centralized along the F2 dimension. Like the vowels produced by speaker FC, it is the case that some movement is present, but each vowel phoneme has different patterns of movement. Next, I consider the vowels of speaker LG, whose vowel plots are shown below in Figures 8 and 9. 109
FIGURE 8: F1 x F2 vowel plot Vowel Plot from Speaker LG, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
In this F1 x F2 vowel plot, like the others, I note that most vowel averages are well separated from each other. The averages for long /üː/ and short /ü/ are very similar, suggesting that the difference between these two vowels is only in duration, without much difference in acoustic quality. Similar situations exist between short /o/ and long /oː/, and, to a somewhat 110
lesser extent, short /u/ and long /uː/. Also note that, for this speaker, long /eː/ begins in a position
(at the 20% timepoint) that is somewhat high and central in the vowel space, before transitioning into a more mid front position, similar to short /e/. This is not an error. A very common suffix on
Kawaiisu verbs is /-kweː/. Several different possible glosses are given for this morpheme by
Zigmond et al. (1990), but the general idea is that this morpheme seems to indicate motion away from a reference point. The apparent glide associated with /eː/ in LG’s vowel space is likely due to the influence of the labiovelar /kw/ preceding the long /eː/, as many of this speaker’s long /eː/ tokens were produced in forms containing the /-kweː/ morpheme. Acoustically, a [w] segment should be quite similar to the vowel [u], having both a low F1 and low F2. In the case of LG, the
20% timepoint average of the /eː/ tokens is both lower in both F1 and F2 than one might expect for a prototypical mid front vowel (and certainly compared to the short /e/ tokens as well as the
50% and 80% averages for long /eː/). For this speaker, because many /eː/ tokens were produced in the /-kweː/ morpheme, the vowel begins in a position more like a [w] or [u] than [e], transitioning into a position more typical for an [e] by the end of the articulation. Below in
Figure 9, a F1 x F3 vowel plot is presented for speaker LG. 111
FIGURE 9: F1 x F3 vowel plot from speaker LG featuring F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
The data from speaker LG were analyzed using the procedures outlined above. Below,
Table 7 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair.
TABLE 7: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker LG, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 103 20.282*** F<1 F<1 /e/ Excluded from analysis due to paucity of data /i/ 1, 143 6.414* 14.515*** F<1 /o/ Excluded from analysis due to paucity of data /u/ Excluded from analysis due to paucity of data /ü/ 1, 74 F<1 F<1 1.766 * p < .05 ** p < .01 *** p < .001 112
Significant differences of vowel quality were found for the /a/ vowel pair and the /i/ vowel pair, which is consistent with the information shown in the vowel plots for speaker LG, seen above. Long /aː/ appears lower in the vowel space than short /a/ (consistent with the significant F1 difference shown in Table 7), while long /iː/ appears both higher and further forward than short /i/, fitting with the significant differences in both F1 and F2 of the /i/ vowel pair. Visually, other long and short pairs appear to be separated in LG’s vowel plots (the /e/ and
/u/ pairs, for instance), but these were not tested for statistical significance because very few tokens of some of these vowels were produced. Below, Table 8 shows F-ratios and significance levels for effect of timepoint on formant structure, reflecting movement in the vowel space.
TABLE 8: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker LG. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 172 12.411*** F<1 3.613* /aː/ 2, 34 9.404*** 3.381* F<1 /e/ Excluded from analysis due to paucity of data /eː/ 2, 40 16.686*** 48.261*** 2.480 /i/ 2, 246 5.472** 24.871*** 5.927** /iː/ 2, 40 1.635 2.157 F<1 /o/ 2, 46 1.602 F<1 F<1 /oː/ Excluded from analysis due to paucity of data /u/ 2, 44 2.321 8.222*** 12.858*** /uː/ Excluded from analysis due to paucity of data /ü/ 2, 134 9.404*** 3.766* 18.256*** /üː/ 2, 14 2.382 1.423 5.143* * p < .05 ** p < .01 *** p < .001
These results show that several of the vowels produced by speaker LG tend to show some degree of movement. This is apparent in the vowel plots above, where it can be seen that both /a/ 113
and /aː/ shift to a slightly lower position in the vowel space over time, and /i/ moves further forward, for example. Of particular interest is long /eː/, which begins very far back in the vowel space (relative to what one may expect from a front vowel), before moving lower and dramatically further forward in the vowel space (as is apparent due to the high F-ratio for F2).
This is expected however, as most of LG’s /eː/ tokens were found in the verbal suffix /-kweː/, as discussed, with heavy influence from the preceding /kw/. Finally, I consider the vowel data from speaker RB. Her vowel plots are shown below in the following figures.
FIGURE 10: F1 x F2 vowel plot Vowel Plot from Speaker RB featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
114
For the most part, this plot seems fairly straightforward, with many of the vowels being reasonably separated in the vowel space. Like LG, the /ee/ tokens produced by RB begin at a lower F2 (at the 20% timepoint) before transitioning into a higher F2 (at the 50% and 80% timepoints). This is not surprising, because like LG, many of RB’s /eː/ tokens were produced within the /-kweː/ suffix. However, some of the results shown here are somewhat unexpected and require explanation. First, note that there are no tokens of /e/ in RB’s data. This is not due to intentional omission, but rather due to the fact that short /e/ is comparatively rare in stressed position in Kawaiisu, and there were simply no tokens of this phoneme that were produced by
RB which fit the criteria for analysis. Second, note the high degree of apparent overlap in the back vowels. The vowels /u/ and /o/ are practically on top of each other, and the long vowels /oː/ and /uː/ vary widely in F1 and F2 depending on timepoint. Additionally, the short /ü/ seems to be encroaching upon the space for the vowel /eː/. The unexpected appearance of back vowels in this plot is likely due to the fact that RB produced only a very small number of back vowels that were suitable for analysis: only one /oː/, only four of /u/, only one /uː/, and only four of /üː/ (recall these totals from Table 2). Short /o/ and /ü/ were slightly better represented. Of the vowels with fewer than 5 tokens, it is probable that if more tokens had been produced that fit the criteria for inclusion in the analyses presented here, the formant averages would even out such that the space for the back vowels would appear more typical, as it tends to be for the other speakers. Below in
Figure 11, a F1 x F3 vowel plot is presented for speaker RB. 115
FIGURE 11: F1 x F3 vowel plot from speaker RB featuring F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
116
The data from speaker RB were analyzed using the same procedures as the data from other speakers. Below, Table 9 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair.
TABLE 9: F-ratios and significance levels comparing long and short vowel quality, for each vowel produced by speaker RB, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 50 8.946** 1.223 F<1 /e/ Excluded from analysis due to paucity of data /i/ 1, 59 F<1 5.695* 7.708** /o/ Excluded from analysis due to paucity of data /u/ Excluded from analysis due to paucity of data /ü/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
In Table 9, only two vowel pairs are tested: the /a, aː/ pair and the /i, iː/ pair. This is due to the fact that RB produced very few (or, in the case of /e/, zero) tokens that fit the criteria for inclusion in the analyses, as seen in Table 2 earlier in this chapter. Significant differences were found in vowel height between the /a, aː/ pair, with long /aː/ being much lower in the vowel space than short /a/, while being very similar in the front-back dimension, as seen in Figure 10. A significant difference was found in F2 for the /i, iː/ pair, with long /iː/ being much further forward in the vowel space than short /i/, while being roughly equal along the F1 dimension. If the /u, uu/ and /ü, üː/ vowel pairs had been included in the analysis, it is possible that significant differences would emerge with these pairs, as Figure 10 shows long /uː/ and long /üː/ to be fairly high in the vowel space compared to their short counterparts; however, including these vowel pairs in the analysis would likely yield results that may be difficult to interpret, due to the relative lack of tokens (for long /üː/, and both members of the /u, uː/ pair). Next, Table 10 shows 117
F-ratios and significance levels for effect of timepoint on formant structure, reflecting some degree of movement.
TABLE 10: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker RB. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 56 4.166* F<1 3.228* /aː/ 2, 44 15.433*** 8.459*** 3.920* /e/ Excluded from analysis due to paucity of data /eː/ 2, 44 F<1 15.032*** 9.377*** /i/ 2, 78 F<1 6.051* F<1 /iː/ 2, 40 F<1 2.209 1.591 /o/ 2, 18 1.266 3.272 F<1 /oː/ Excluded from analysis due to paucity of data /u/ Excluded from analysis due to paucity of data /uː/ Excluded from analysis due to paucity of data /ü/ 2, 66 F<1 F<1 1.528 /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
In Table 10, it can be seen that some significant differences are found between timepoints with some of the vowel phonemes, indicating some movement in /a/, /aː/, /eː/, and /i/. This is consistent with the vowel plots shown above. The short /a/ vowel drops slightly in the vowel space, while long /aː/ drops and moves slightly forward throughout its articulation on average.
Short /i/ remains at roughly the same height, but does move forward somewhat.
2.4 Discussion In this chapter, the stressed vowel system of the Kawaiisu language was examined.
Zigmond et al. (1990) make the claim that in Kawaiisu, vowels can be either long or short, with the primary difference between long and short vowels being duration. This claim has been 118
confirmed in this chapter. A statistically significant difference was found in duration between the vowels Zigmond et al. (1990) claim to be short, and those that are claimed to be long, with the phonemically long vowels having a much greater duration on average.
Additionally, the results of the analyses presented in this chapter show that the long vowels often, but not always, occupy a more peripheral position in the vowel space: the long vowels usually appear closer to the edges of the vowel space, with the short vowels pushed more towards the center. Although, due to scarcity of data in some cases, there were some vowels which could not be included in the statistical analyses which compared the formant structure of long and short vowel pairs. However, even in these situations it is often apparent that the long vowels lean more towards the periphery of the vowel space. For example, speaker AP’s /u/, /uː/ and /o/, /oː/ vowel pairs were not compared in this way because so few tokens were produced, and such analyses would not have much statistical power (leading to a greater potential for Type
II error), but these vowel pairs nevertheless follow the same pattern: the long member of the pair tends to be closer to the extreme edges of the speaker’s vowel space, while the short member of the pair tends to fall closer to the center. Other speakers show similar results, even in some cases where the vowels were not tested due to paucity of data.
One robust exception to this tendency can be found in the /e/, /eː/ vowel pair. For this pair, the short /e/ often appears with a higher F2 in a more front position in the vowel space, while /eː/ seems to lie more towards the center. The most plausible explanation for the apparent centralization of /eː/ in the F1 x F2 vowel plots presented in this chapter is that many of the /eː/ tokens selected for analysis are part of the /-kweː/ verbal suffix, and show heavy influence from 119
the preceding /kw/ as the formants glide from those associated with a labiovelar articulation to a more forward place of articulation for the vowel.
Additionally, many of the vowels were found to have at least some degree of movement throughout their articulation. Within speakers, each vowel tends to have a different pattern of movement, with some vowels rising, others falling, some moving forward, and others moving further back in the vowel space. Between speakers, however, a few consistent patterns can be found. For most of the speakers where statistically significant differences were found, short /a/ tends to fall in the vowel space (although it rises slightly in FC’s data). Long /aː/ tends to lower, then rise slightly for all speakers (including in AP’s data, although this difference was not statistically significant). Long /ee/ moves forward in the vowel space for all speakers. This forward movement, associated with an increase in F2, is likely due to the influence of the /-kweː/ morpheme, as previously discussed. Also, there is a general tendency for this vowel to become slightly heightened, although the height difference in F1 was not statistically significant for all speakers. Short /i/ tends to rise slightly in the vowel space and move further forward for all speakers, becoming more peripheral throughout the vowel’s articulation. Other vowels did not display the same consistent patterns across speakers, with /o/, for instance, rising for some speakers and lowering for others. The vowel /ü/ moves further back in the vowel space for some speakers, and further forward for others. Overall, however, the data suggest that some uniformity across speakers exists for several of the vowels that were tested. 120
CHAPTER 3: UNSTRESSED VOWELS
3.1. Introduction In this chapter, I examine the unstressed Kawaiisu vowel space. Chapter 2 presented an outline of the stressed vowels, showing that previous research (Klein 1959, 2002, Zigmond et al.
1990, Thomas 2017) was essentially correct in identifying six phonemic vowel qualities: /i, e, u,
ü, a/. In stressed position, all six of the Kawaiisu vowels can appear as either long or short. The distinction between long and short vowels in Kawaiisu is phonemic, and does not arise due to phonological alternation (although some long vowels are the result of morphological concatenation and vowel coalescence, as discussed in Chapters 1 and 2).
Outside of stressed syllables, the distinction between long and short vowels still occurs.
Such examples are numerous in Zigmond et al. (1990). Consider the examples /naɣaˈvivi/
/naːˈkeːdü/
For the most part, stressed and unstressed syllables seem to share similar phonological characteristics: all six Kawaiisu vowels can occur in either syllable type, and they can be either long or short. However, in syllables which do not carry stress, vowels are often deleted, especially word-finally. Word-internally, this type of deletion is common in English. For example, many speakers may not pronounce the vowel in the second syllable of a word like family , which is unstressed, preferring instead a pronunciation more like [ˈfæm.li]. The phenomenon of word-internal unstressed vowel deletion is not unique to English or Kawaiisu, and has been reported to occur in a number of languages (see for instance Hooper 1978, Silva
1997, Johnson 2004, Delforge 2008a, Torreira & Ernestus 2011, Schuppler et al. 2011, etc. ). In this chapter, I discuss the phenomenon of word-internal vowel deletion in Kawaiisu, noting the phonological environments where such deletion tends to occur most often.
This process of word-final deletion is apparently optional, with Zigmond et al. (1990) noting that it tends to occur utterance-finally, and particularly in verbs. Vowel devoicing is not an unusual phenomenon in the Numic family, as noted by Sapir (1931), Charney (1993), and others. However, Zigmond et al. (1990) point out that the process of vowel deletion in Kawaiisu is unusual within Numic as it tends to affect final vowels only. Aside from this, similar processes of vowel devoicing in other Numic languages may be a product of predictable phonological alternation (as Charney 1993 discusses for some instances of Comanche vowel devoicing), but the process of deleting word-final vowels in Kawaiisu is apparently entirely optional, and does not seem to be governed in a predictable way by phonological rule. In addition to analyses of word-internal vowel quality, I also examine the process of word-final vowel deletion.
122
3.2. Methods The methods for this chapter are similar to the methods used in the previous chapter on the stressed vowel space of Kawaiisu. The data are from Sheldon Klein’s field recordings (Klein
1958, 1981-1983), and the words selected for analysis come from his elicitation sessions, rather than from stories, monologues, or conversational speech. These words were produced in isolation.
Vowels selected for durational and spectral analysis in this chapter were produced in unstressed, non-final syllables. Word-final vowels, although usually unstressed, were not included in the analyses in order to negate the effect that word-finality may have on duration (as discussed in Section 2.2 of this dissertation). Vowels in penultimate syllables were not considered for analysis in this chapter, as this is the usual position for primary lexical stress in
Kawaiisu. Thus, all tokens selected for durational and spectral analysis in this chapter may come from any syllable preceding the penultimate.
The criteria for selecting the beginning and end points of unstressed vowels, and the phonological environments used, were the same as in the previous chapter, summarized briefly here. Vowels in pre-nasal environments were not selected for analysis. Vowels which do appear in these analyses may come from any other phonological environment (after a nasal consonant, or bounded on either side by either an obstruent or glide). Onset and offset of F2 were used to determine the vocalic boundaries when the vowel appeared in an obstruent-containing environment. In cases where F2 remains weakly present during the articulation of a consonant, the boundary is taken to be the point where F2 most suddenly gains or loses amplitude. In post- nasal environments, the beginning of the vowel is signaled by a sudden shift of acoustic energy in the spectrum as the nasal transitions into a vowel articulation. In glide-containing 123
environments, boundaries were placed at the specific point in the waveform where a sudden change in amplitude is apparent, roughly halfway between the formant transitions between the glide and the actual vowel. After all relevant vowel boundaries were placed, Praat (Boersma &
Weenink 2017) was used to extract the specific duration of each vowel token, in addition to measurements of F1, F2, and F3 at timepoints of 20%, 50%, and 80% for each vowel. As in the previous chapter, surrounding consonantal environment could not be tightly controlled, and any effect that adjacent consonants may have had on vowel duration could not be measured. Finally, when determining whether to consider a vowel as long or short, I refer to the Zigmond et al.
(1990) dictionary: if they list a given vowel as long, I consider it to be long, and if they list a particular vowel as short, I consider it to be short for the purposes of these analyses.
3.3. Durational and Segmental Results 3.3.1 Duration Analyses
Using the methods discussed above, a total of 1,864 vowel tokens were analyzed. A table showing the number of unstressed vowel tokens from each speaker is shown below in Table 1, followed by Figure 1, which shows the difference in duration between phonemically long and phonemically short unstressed vowel pairs. 124
TABLE 1: Number of unstressed vowel tokens available for analysis for each speaker and vowel phoneme (long and short vowels listed separately). This represents the number of each unstressed vowel in the phonological environments described for inclusion above. Speaker Vowel phoneme a aː e eː i iː o oː u uː ü üː AP 131 21 7 47 71 19 64 8 38 1 43 4 FC 72 9 9 13 25 2 45 1 36 5 21 1 LG 300 30 17 20 165 6 142 15 97 22 97 6 RB 67 20 4 15 39 4 35 5 24 3 37 1 Total 570 80 37 95 300 31 286 29 195 31 198 12
FIGURE 1: Comparison of long and short vowel durations in unstressed position, averaged over all speakers and tokens
125
In Figure 1, above, each bar represents the mean duration for each unstressed vowel, averaged across all 4 speakers. For each pair, the phonemically short vowels, represented in blue, have a shorter duration than the phonemically long vowels, represented in green. As shown in
Figure 1, short unstressed vowels tend to have an average around 50-60ms, while long vowels tend to average around 100-150ms. For some vowel pairs, the difference in duration between short and long was more drastic than other vowel pairs; for instance, short /i/ averages 57ms, while long /iː/ averages 101ms, a difference of 44ms. Short /ü/, on the other hand, averages only
48ms (the shortest average duration of the short vowels), while long /üː/ shows an average of
154ms (the longest average duration of the long vowels), attaining a difference of over 100ms.
Investigating the difference in duration between phonemically long and short unstressed vowels, I conducted a two-factor within-subjects ANOVA, with data averaged over words in each condition for each speaker. In this test, the two factors were phonemic length (with separate levels for long and short vowels), and vowel quality (one level for each vowel quality, for a total of six levels of the factor). There was a significant main effect of phonemic length ( F(1, 3) =
114.613, p = 0.002), showing that there is indeed a significant difference in duration between phonemically long vowels and phonemically short vowels in unstressed position. The main effect of vowel quality was significant ( F(5, 15) = 6.018, p = 0.003), and the interaction of the two factors was significant as well ( F(5, 15) = 6.705, p = 0.002), indicating that the variation in duration between vowel pairs differs depending on vowel quality, as illustrated in the above discussion of the difference in duration between the /i/, /iː/ pair and the /ü/, /üː/ pair.
Examining this interaction further, I conducted a series of pairwise comparisons of vowel qualities with the same phonemic length. For the short vowels, significant differences were 126
found between /a/ and /ü/ ( t(3) = 29.128, p < 0.001), /e/ and /ü/ ( t(3) = 3.227, p = 0.048), /i/ and
/ü/ ( t(3) = 5.711, p = 0.011), /o/ and /ü/ ( t(3) = 4.279, p = 0.023), and /u/ and /ü/ ( t(3) = 3.261, p =
0.047). These findings fit well with the previously discussed fact that short /ü/ tends to have the shortest duration of all the vowels, averaging less than 50ms.
These comparisons were also conducted on the long vowels. In these tests, significant differences were found between /aː/ and /iː/ ( t(3) = 5.227, p = 0.014), /eː/ and /iː/ ( t(3)= 4.380, p
= 0.022), /eː/ and /üː/ ( t(3) = -3.482, p = 0.040), /iː/ and /üː/ ( t(3) = -4.313, p = 0.023), /oː/ and
/uː/ ( t(3) = 4.223, p = 0.024), and /uː/ and /üː/ ( t(3) = -4.419, p = 0.022). This is consistent with the general pattern, which can be seen above in Figure 1, which suggests greater durational variation in the phonemically long vowels.
3.3.2 Segmental Results
The results of the spectral analyses of unstressed vowels are presented in a similar way to the spectral analyses of stressed vowels seen in the previous chapter of this dissertation. Each speaker’s vowels were analyzed separately, and results from each of the four speakers will be presented in turn. Vowel plots will be shown for each speaker (F1 x F2 and F1 x F3), and any visually apparent patterns in the plots will be discussed.
Each speaker’s vowel data were analyzed separately using a series of statistical tests. In the first group of tests, by-items ANOVAs were conducted on each speaker’s long and short vowel pairs, testing for differences of vowel quality (/i/ is compared to /iː/, /o/ to /oː/, etc. ). In this group of tests, the dependent variables are F1, F2, and F3. The script used to extract formant information collected data from three separate timepoints for each vowel token (20%, 50%, and 127
80%), but in this particular group of tests, F1, F2, and F3 are averaged across these three timepoints, such that each vowel token has a single value for each formant.
The second series of tests looks at the formant values of each vowel token, at each timepoint, to see how each vowel moves in the vowel space throughout its articulation. A high degree of variation in formants at different points in time could indicate some movement in the vowel space. These tests were repeated measures ANOVAs, with timepoint as a within-subjects factor (with three levels, one for each timepoint), and F1, F2, and F3 as dependent variables.
Below, Figure 2 shows the F1 x F2 for the unstressed vowels of speaker FC.
FIGURE 2: F1 x F2 Vowel Plot from Speaker FC, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
128
In speaker FC’s unstressed F1 x F2 plot, there is some slight crowding in the mid to high range among the front vowels (and short /ü/), but for the most part, these vowels seem to be relatively well-separated in the vowel space. Long vowels occur closer to the edges of the vowel space, while short vowels tend to have more centralized averages. An F1 x F3 vowel chart of
FC’s unstressed vowels is shown below, in Figure 3.
FIGURE 3: F1 x F3 Vowel Plot from Speaker FC, with F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
These vowel data seen in Figures 2 and 3 were analyzed using the tests described above in this section. Below, Table 2 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair. 129
TABLE 2: F-ratios and significance levels comparing long and short vowel quality, for each vowel, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 80 12.967** F<1 F<1 /e/ 1, 21 2.356 27.113*** F<1 /i/ Excluded from analysis due to paucity of data /o/ Excluded from analysis due to paucity of data /u/ 1, 40 F<1 3.101 1.099 /ü/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
Due to the fact that FC produced only a few tokens of some vowels (see Table 1 for a token count for each speaker), some vowel pairs were excluded from these analyses: the /i/, /iː/ pair, the /o/, /oː/ pair, and the /ü/, /üː/ pair. Of the vowels which were included in the analyses, only two statistically significant differences were found. F1 of unstressed /a/ and /aː/ differs significantly, which is consistent with the vowel plots above, where it is seen that /aː/ has a noticeably higher F1 than /a/, and as such, lies lower in the vowel space. Additionally, F2 differs significantly for the /e/, /eː/ pair. Note that, in the vowel plot above, these vowels seem to be similar in height, but /eː/ appears further forward, approaching the periphery, while /e/ appears in a more central position. This pattern is apparent even for those vowels which were not included in the analyses: long vowels are more peripheral in the F1 x F2 vowel space, while short vowels appear more centralized. Below, Table 3 shows F-ratios and significance levels for effect of timepoint on formant structure. 130
TABLE 3: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker FC. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 142 5.380** F<1 15.630*** /aː/ 2, 16 1.340 F<1 13.069*** /e/ 2, 16 9.900** 1.543 F<1 /eː/ 2, 24 10.581** F<1 1.649 /i/ 2, 48 4.207* 2.303 F<1 /iː/ Excluded from analysis due to paucity of data /o/ 2, 88 2.888 F<1 5.137** /oː/ Excluded from analysis due to paucity of data /u/ 2, 70 1.415 F<1 7.854** /uː/ 2, 8 2.940 1.385 1.222 /ü/ 2, 40 F<1 8.332** 3.383* /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
The tests above in Table 3 show that some significant differences can be found in vowel quality among the three different timepoints for some of the vowels. Statistically significant differences in F1 were found for /a/, /e/, /eː/, and /i/. These vowels tend to rise in the vowel space throughout their articulation. A significant difference in F2 was found for the vowel /ü/, which is shown to move further back in the vowel space. F3 was significantly different for FC’s /a/, /aː/,
/o/, /u/, and /ü/ vowels, with F3 decreasing over time for /a/, /aː/, and /ü/ (all unround vowels), but increasing over time for /o/ and /u/, both round vowels. Next, consider the data for speaker
AP. Below, an F1 x F2 plot of her unstressed vowels is shown in Figure 4. 131
FIGURE 4: F1 x F2 Vowel Plot from Speaker AP, with F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
For the most part, AP’s unstressed vowel averages seem reasonably well-separated.
Many of her long and short vowel pairs appear to occupy distinct areas in the vowel space. These vowel pairs have been tested for statistically significant differences in vowel quality, and the results of these tests are shown below in Table 4. The /u/, /uː/ pair and the /ü/, /üː/ pair have been omitted from these tests due to an inadequate number of tokens, but even these vowels show the same general trend: long vowels are closer to the edge, while short vowels are more centralized.
One exception can be found in the /e/, /eː/ pair. A visual examination of the plot above suggests that this vowel pair is not likely to be qualitatively distinct; and indeed, no statistically significant 132
differences in quality are found for this vowel pair, as shown below in Table 4. Below, Figure 5 shows an F1 x F3 vowel plot for speaker AP’s unstressed vowels.
FIGURE 5: F1 x F3 Vowel Plot from Speaker AP, with F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
In this plot, it is shown that speaker AP’s unstressed vowels tend to fall in an F3 range of around 2700-2900Hz, with a few outliers: /uː/ is shown to have a much higher F3, /oː/ (at least at timepoints 50% and 80%) has a much lower F3. /a/ and /aː/ are closer to the F3 cluster, but tend to have F3 averages that are generally a little lower than the majority of vowels. 133
AP’s vowel data were analyzed using the statistical tests outlined above. Below, Table 4 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair.
TABLE 4: F-ratios and significance levels comparing long and short vowel quality, for each vowel, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 150 36.204*** F<1 1.129 /e/ 1, 53 F<1 F<1 F<1 /i/ 1, 89 F<1 19.361*** 2.944 /o/ 1, 70 8.507** 10.488** 3.960* /u/ Excluded from analysis due to paucity of data /ü/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
The results presented above in Table 4 show that there are some significant qualitative differences between long and short unstressed vowel pairs. Long /aː/ is much lower in the vowel space than short /a/, closer to the periphery. This is consistent with the results presented above, which shows that this vowel pair has a highly significant difference in F1 (p < 0.001). Short /i/ and long /iː/ are very similar in vowel height, but long /iː/ is much further forward, closer to the edge of the vowel space, consistent with the results presented here showing a highly significant difference in F2 for this vowel pair (p < 0.001). Significant differences were found for all 3 formants of the /o/, /oː/ pair, consistent with the information shown in the vowel plots above, where it can be seen that long /oː/ is both higher and further back than short /o/, in addition to having lower F3 on average. Below, Table 5 shows F-ratios and significance levels for effect of timepoint on formant structure. 134
TABLE 5: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker AP. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 260 3.331* 3.360* F<1 /aː/ 2, 40 1.556 2.766 1.142 /e/ 2, 12 10.516** F<1 1.034 /eː/ 2, 92 6.957** 1.201 1.415 /i/ 2, 140 8.622*** 2.593 F<1 /iː/ 2, 36 1.011 2.720 2.846 /o/ 2, 126 3.537* 2.552 F<1 /oː/ 2, 14 10.224** 2.804 5.379* /u/ 2, 74 5.562** F<1 1.191 /uː/ Excluded from analysis due to paucity of data /ü/ 2, 84 1.621 F<1 F<1 /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
In AP’s data, some patterns were found for vowels where the formant structure was found to vary significantly across timepoint. Where statistically significant differences are found, they tend to be in F1, affecting 7 of the 12 vowel phonemes. Generally, it is the case that AP’s vowels rise slightly in the vowel space over time. This is true for /e/, /eː/, /i/, /u/, and /oː/. A significant difference in F3 was also found for /oː/, which is shown to decrease in F3 over time, as seen in Figure 5 above.
Significant differences across timepoint were also found in F1 of /a/ and /o/ (/a/ also differs in F2). However, an examination of the relevant vowel plots indicates a relatively small variance of formant structure across the three timepoints. That is, the averages for these vowels at all three timepoints appear to be very close together. Compared to the other vowels produced by AP, /a/ and /o/ have a fairly large number of tokens represented in the data (as indicated in the 135
token count shown in Table 1, and the degrees of freedom shown in Table 5). Only /i/ has a similarly high token count in AP’s data. These findings suggest that, although the variation in formant structure for /a/ and /o/ may be small, it must be consistent across the majority of tokens produced for each of the vowels, in order for statistical significance to be attained. Next, I consider the unstressed vowels of speaker LG, whose F1 x F2 unstressed vowel plot is shown below in Figure 6.
FIGURE 6: F1 x F2 Vowel Plot from Speaker LG, featuring F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
In this vowel plot, long /iː/ is clearly distinct from short /i/, with the long vowel being much further forward, while the short vowel is somewhat centralized. The /a/, /aː/, and /o/, /oː/ pairs appear to differ in vowel height, but not along the front-back dimension, while the /u/, /uː/ 136
pair seems to differ along the front-back dimension (with the long vowel closer to the edge of the vowel space), but not in height. The /ü/, /üː/ pair does not seem to show any difference in vowel quality, with the long and short members of the pair clustered together in the plot. Below, an F1 x F3 vowel plot for speaker LG appears in Figure 7.
FIGURE 7: F1 x F3 Vowel Plot from Speaker LG, with F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
In this plot, it can be seen that many of the vowels produced by LG have an F3 average in the 2700-2800Hz range, similar to the F1 x F3 plot shown above for speaker AP, whose vowel
F3 averages tended to fall in a similar range. A couple exceptions are obvious: long /iː/, which tends to have a higher F3 than the majority of LG’s vowels, and the /u/, /uː/, and /oː/ vowels which tend to have a lower F3 than most vowels in the plot. Recalling that the round vowel 137
phonemes of Kawaiisu are /u/, /uː/, /o/, /oː/, it is not surprising that these vowels should have a lower F3, as low F3 is often associated with rounding of vowels.
Results of the statistical analyses for LG’s unstressed vowels are shown below. First,
Table 6 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair.
TABLE 6: F-ratios and significance levels comparing long and short vowel quality, for each of LG’s unstressed vowels, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 329 12.766*** F<1 F<1 /e/ 1, 36 12.646*** F<1 F<1 /i/ 1, 170 9.250** 13.845*** 5.201* /o/ 1, 155 5.026* F<1 F<1 /u/ 1, 118 1.856 1.169 F<1 /ü/ 1, 102 F<1 F<1 4.463* * p < .05 ** p < .01 *** p < .001
The table above shows that the majority of differences are along the height dimension.
Where qualitative differences exist in LG’s unstressed vowel pairs, they differ in height more often than backness. The only unstressed vowel pair where backness is significantly different is the /i/, /iː/ pair, in which a highly significant difference was found for F2. Given the plot shown about in Figure 6, this is not surprising, as it can be seen that long /iː/ approaches the extreme front edge of the vowel space, with a very high F2, while short /i/ is more centralized (often with lower F2 averages than /e/). Below, Table 7 shows F-ratios and significance levels for effect of timepoint on formant structure. 138
TABLE 7: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker LG. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 598 10.622*** 17.385*** 2.564 /aː/ 2, 58 5.303** F<1 F<1 /e/ 2, 32 F<1 F<1 F<1 /eː/ 2, 38 2.102 10.136*** F<1 /i/ 2, 328 21.695*** 4.216* F<1 /iː/ 2, 10 F<1 F<1 10.007** /o/ 2, 282 8.953*** 1.121 3.062* /oː/ 2, 28 F<1 F<1 3.869* /u/ 2, 192 11.154*** 10.977*** F<1 /uː/ 2, 42 F<1 6.556** 2.448 /ü/ 2, 192 6.984** 7.930*** F<1 /üː/ 2, 10 F<1 4.078 1.025 * p < .05 ** p < .01 *** p < .001
In the vowel plots presented above for speaker LG, it is apparent that some vowels show a high degree of variation between different timepoints. This is especially apparent for the long vowels /aː/ and /eː/. The vowel plot above shows that F1 /aː/ tends to rise approximately 50Hz on average, before lowering roughly 100Hz on average, thus falling, then rising, in the vowel space.
Long /eː/, on the other hand, shows an increase in F2 of approximately 400Hz, moving throughout the duration of articulation from a more centralized position to one that is slightly more front. As discussed in Chapter 2, this fronting of long /eː/ may be due to the high frequency verbal suffix /-kweː/, which often occurs in both stressed and unstressed positions.
Some of the results presented above in Table 7 correspond with variation that is less obviously apparent based on the vowel plots presented above in Figures 6 and 7. For example,
Table 7 reports a statistically significant difference in F1 across the three timepoints for the short 139
vowel /i/ (p < 0.001). However, examining Figure 7 shows that the average difference in F1 for the short vowel /i/ between the three timepoints is less than 50Hz. This is not necessarily a large difference, but it is likely to be a consistent one. Similar issues were seen in AP’s unstressed vowel variation, likely due to a very high token count for certain vowel phonemes. Finally, the unstressed vowels of speaker RB are considered, whose vowel plots are presented below in
Figures 8 and 9.
FIGURE 8: F1 x F2 Vowel Plot from Speaker RB, with F1 in Hz on the vertical axis, and F2 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
In this F1 x F2 vowel plot featuring the unstressed vowels of speaker RB, it appears that long /iː/ tends to vary widely by timepoint, but remains reasonably separated from short /i/. Long
/uː/ also displays a high degree of variation across the three timepoints. Short /ü/ appears vary far 140
forward in the vowel space, essentially collapsing with /e/ and /eː/. As shown above in Table 1, the number of unstressed vowel tokens produced by this speaker was comparatively small. As the visual information in vowel plots, like Figure 8, is based on average formant values, it is possible that the high degree of variation and seemingly aberrant behavior of some of the vowels in this plot is due to the averaging together of a very small number of tokens.
FIGURE 9: F1 x F3 Vowel Plot from Speaker RB, with F1 in Hz on the vertical axis, and F3 in Hz on the horizontal axis. Long vowels are represented by triangles, and short vowels by squares. Blue, red, and black markers represent timepoints 20%, 50%, and 80% respectively.
The unstressed vowel data collected from speaker RB and represented in Figures 8 and 9 were subjected to the same statistical analyses outlined above for the other speakers. Results of 141
these analyses are presented below in Tables 8 and 9. First, Table 8 details F-ratios and significance for effects of phonemic length on vowel quality for each long and short vowel pair.
TABLE 8: F-ratios and significance levels comparing long and short vowel quality, for each vowel, with F1, F2, and F3 averaged across timepoints. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 1, 86 1.441 F<1 F<1 /e/ Excluded from analysis due to paucity of data /i/ Excluded from analysis due to paucity of data /o/ 1, 39 F<1 F<1 F<1 /u/ Excluded from analysis due to paucity of data /ü/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
Table 8 shows that none of the vowel pairs tested were significantly different in quality.
Four vowel qualities (/e/, /i/, /u/, /ü/) were excluded from the analyses because, for some phonemes, RB simply did not produce enough unstressed vowel tokens which fit the criteria for analysis. Of the vowel pairs which were tested, the results shown in Table 8 are in accordance with the information presented in Figures 8 and 9. Both figures show that short /o/ and long /oː/ are clustered tightly together. Figure 8 shows that long /aː/ tends to fall slightly lower in the vowel space than short /a/ (which fits the general trend that long vowels occur closer to the periphery), but this difference was not statistically significant.
Some of the other vowel pairs which were not tested due to an inadequate number of tokens appear very distinct in the above plots. These include the vowel pairs /i/, /iː/, and /ü/, /üː/.
For each of these pairs, a considerable number of short vowel tokens were collected, but the token count for the long members of these pairs was quite small. If additional long vowel tokens 142
which fit the criteria for analysis had been produced by this speaker, it is possible that the analyses would have obtained statistically significant differences for these vowel pairs, given their apparently distinct averages in the vowel plots shown in Figures 8 and 9. However, even if only one of the long vowel tokens collected for these pairs was not typical of the phoneme, this could have a large effect on the formant averages, and by extension, the graphical representations of those averages seen in the vowel plots in Figures 8 and 9. Thus, even though these long and short vowel pairs appear distinct in the plots above, the available data are not sufficient to draw any strong conclusions about any qualitative differences (or lack thereof) between members of these vowel pairs. Below, Table 9 shows F-ratios and significance levels for effect of timepoint on formant structure.
TABLE 9: F-ratios and significance levels comparing differences in timepoint (20%, 50%, 80%) on F1-F3 for each vowel phoneme for speaker RB. Significance is marked as indicated, and appears in bold. Vowel type Degrees of freedom F1 F2 F3 /a/ 2, 132 2.546 9.029*** F<1 /aː/ 2, 38 10.021*** 1.269 F<1 /e/ Excluded from analysis due to paucity of data /eː/ 2, 28 1.149 5.315* 1.536 /i/ 2, 76 4.394* F<1 F<1 /iː/ Excluded from analysis due to paucity of data /o/ 2, 68 12.121*** F<1 F<1 /oː/ 2, 8 F<1 F<1 F<1 /u/ 2, 46 F<1 F<1 1.611 /uː/ Excluded from analysis due to paucity of data /ü/ 2, 72 3.759* 1.773 2.222 /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
143
For each of RB’s vowels, F3 remains fairly consistent throughout their articulation. None of the vowels that were tested differ significantly in F3 across timepoints, as can be seen above in Table 9. Many of the vowels, however, vary in either height or backness. None of RB’s unstressed vowels vary significantly along multiple dimensions. For the vowels that vary along
F1, most show a pattern of rising. These include /i/, /o/, and /ü/. The only vowel which differs in
F1 and falls in the vowel space is long /aː/. Along the F2 dimension, short /a/ moves further back in the vowel space, while long /eː/ is somewhat fronted.
Overall, the results presented in this section show some qualitative differences between long and short vowels for each speaker, and some (often small) differences between different timepoints for different vowel phonemes for each speaker. While there no statistically significant differences in quality between long and short vowel pairs for speaker RB, there were some consistent patterns with the other speakers. For each of the other speakers, long /aː/ is lower in the vowel space than short /a/. For AP and LG, long /iː/ is further forward in the vowel space than short /i/. The vowel plots for RB and FC show similar trends for this vowel pair, but there were not enough tokens of long /iː/ produced by either of these speakers to allow statistical analysis. Generally, the vowel plots shown in this section illustrate that long unstressed vowels tend to lie closer to the periphery of the vowel space than short unstressed vowels.
Regarding differences across timepoints, short /a/ displays a pattern wherein it falls slightly, then rises. This effect appears small, but was statistically significant for three out of four speakers. Long /eː/ tends to either rise in the vowel space, move forward, or both, for all speakers. As discussed in previous chapters, this is likely due to the relatively high occurrence of the /-kweː/ suffix. Short /i/ tends to rise over time for all four speakers. Generally, when 144
examining the quality of unstressed Kawaiisu vowels at different timepoints, it is usually the case that vowels tend to rise in the vowel space, rather than fall. This trend can be seen across all four speakers.
3.4. On Vowel Deletion 3.4.1 Word-Internal Deletion
Numerous studies exist on the phenomenon of the reduction and deletion of vowels in unstressed syllables in many languages (Hooper 1978, Silva 1997, Johnson 2004, Delforge
2008a, Delforge 2008b, Torreira & Ernestus 2011, Schuppler et al. 2011, etc. ). Like other languages, deletion of unstressed vowels often occurs in Kawaiisu as well. Although this is more common in speech at a conversational or rapid pace, the speakers in these recordings (Klein
1958, 1981-1983) occasionally drop vowels from unstressed syllables even in the more careful speech recorded in wordlist elicitation sessions. These vowels tend to be dropped fully, rather than simply devoiced. Vowels were judged to be deleted when there were no visible formants present between two consonants where a phonological vowel would normally be expected, based on the lexicon provided by Zigmond et al. (1990).
When speakers of Kawaiisu delete vowels from unstressed syllables, it is not due to some obligatory phonological rule. Speakers are often heard pronouncing the same word in different ways: sometimes an unstressed vowel is deleted, and sometimes it is present. Below, Figures 10 and 11 show spectrograms and waveforms of speaker LG pronouncing the word /tsiɣüˈpizi/
FIGURE 10: Waveform and spectrogram of /tsiɣüˈpizi/
146
FIGURE 11: Waveform and spectrogram of /tsiɣüˈpizi/
In these figures, the waveform is at the top, followed by a spectrogram. Beneath the spectrogram appear two Praat textgrids, the first labeling each segment that is present in the utterance, the second labeling the phonological word, /tsiɣüˈpizi/
/i/ is absent. This is an example of the similar phenomenon of word-final unstressed vowel deletion, which I return to in Section 3.4.2. More relevant to the current discussion are the word- medial vowels. In Figure 10, all word-medial vowels can clearly be seen in the spectrogram. In the waveform, this is evident as an increase in amplitude, and in the spectrogram, the presence of these vowels is shown by the presence of vowel formants, which are quite visible. In Figure 11, however, the vowel /ü/ is absent. In this utterance, no vowel is present between the [g] and the 147
[p]. A stop burst can be seen as the [g] is released before the [p] closure, but there is no vowel of any substantial or measurable duration between these two segments in Figure 11, as there was in
Figure 10.
This comparison and discussion of Figures 10 and 11 shows that the process of word- internal vowel deletion is not obligatory. The examples shown above in Figures 10 and 11 are not anomalous in Kawaiisu. It is not simply the case that the pronunciation in Figure 11 is a speech error on the part of LG. Speakers may pronounce or omit certain unstressed vowels. It is likely that their presence or absence is tied to rate of speech, as it is in other languages, but a comparison of different speech rates, styles, or registers was not possible within the scope of the current study. A discussion of some factors which may influence vowel deletion in a language is given in Ogasawara (2007).
Certain vowel phonemes were deleted more often than others. In the data that were examined for this chapter, there were a total of 149 instances of word-internal vowel deletion.
The most commonly deleted vowel phoneme was /i/, with 48 deletions, followed by /a/ (37 deletions), /ü/ (34 deletions), /u/ (19 deletions), and /o/ (11 deletions). Neither short /e/ nor any of the long vowels were ever deleted in any of the data examined here. Word-internal deletion affected short vowels only.
Most instances of word-internal unstressed vowel deletion were limited to certain phonological environments where at least one adjacent consonant was voiceless. 62.4% of these deletions occurred between two voiceless consonants, while 18.1% occurred between a voiced consonant on the left and a voiceless consonant on the right, and 5.4% of deletions occurred between a voiceless consonant on the left and a voiced consonant on the right. Thus, 148
approximately 85.9% of word-internal unstressed vowel deletion noted in Klein’s data is associated with voicelessness of an adjacent consonant. Only 14.1% of these deletions occurred between two voiced consonants.
The discussion presented in this section can be summarized as follows. Word-internally, unstressed vowels in Kawaiisu may be deleted. 10 When vowel deletion does occur, it is not obligatory, nor does it categorically affect any specific vowel or class of vowels in any specific phonological environment. General tendencies do exist, however, and the vowels which are most commonly affected by this process are high vowels (especially /i/), which lie between two voiceless consonants (or are, at least, immediately followed by one). Cross-linguistically, it is not unusual for vowel reduction processes to work in this way. Delforge (2008b) discusses a similar situation in Andean Spanish in which reduction is often triggered by the adjacency of voiceless consonants. Beckman (1994) discusses phenomena of vowel reduction, elision, and devoicing in
English, German, Japanese, Korean, and French, working within the framework of Articulatory
Phonology to offer an explanation of such systems. Thus, the phenomenon of unstressed vowel deletion in Kawaiisu appears to be quite similar to other vowel reduction processes found elsewhere in the languages of the world.
10 Vowels may be reduced without being deleted entirely. In English, for example, it is well known that unstressed vowels are often reduced to schwa, and that, cross-linguistically, centralization in the vowel space is common for unstressed vowels. In Chapter 4 of this dissertation, comparisons are made between the stressed and unstressed vowel spaces of Kawaiisu, showing that unstressed vowels in Kawaiisu tend towards centralization as well. However, I have not attempted to quantify potentially varying degrees of vowel reduction here; for the present purpose, vowels are either present or deleted. 149
3.4.2 Word-Final Deletion
As discussed, unstressed vowels in word-medial or word-internal position are often dropped in Kawaiisu speech. Word-final vowels are usually unstressed in Kawaiisu as well, and are often deleted. However, there are some important differences between word-internal and word-final vowel deletion. First, speakers may be largely unaware of the type of vowel reduction/deletion discussed above in Section 3.4.1, but deletion of word-final vowels is salient enough to Kawaiisu speakers that there are orthographic conventions associated with it. In written Kawaiisu, final vowels are often enclosed in parentheses to indicate that sometimes the vowel is pronounced, and sometimes not (Laura Grant, personal communication). Below,
Figures 12 and 13 highlight this optionality, with waveforms and spectrograms of the same speaker producing the same word, once with the final vowel present, and once with the final vowel deleted. 150
FIGURE 12: Waveform and spectrogram of /seːˈɣidü/
Above, Figure 12 shows speaker AP producing the word /seːˈɣidü/
FIGURE 13: Waveform and spectrogram of /seːˈɣidü/
In Figures 12 and 13, the same word is pronounced by the same speaker. In the waveform and spectrogram in Figure 12, evidence of a fully voiced word-final vowel was present. There were visible formants in the spectrogram and an increase in amplitude of the waveform. In
Figure 13, no such evidence exists. There is a very clear stop burst signaling the release of the
[d], but there is nothing in the waveform or spectrogram suggesting the presence of any kind of speech segment afterward. In Figure 13, the word-final vowel has been deleted. Comparing
Figures 12 and 13 (the same speaker producing two separate utterances of the same word) serves to show that both options are possible in Kawaiisu, and similar examples of words being pronounced with and without final vowels can be found in the Klein recordings (Klein 1958,
1981-1983).
Orthographically, word-final vowels can be enclosed in parentheses to highlight the fact that sometimes they are not heard, as noted. Although neither word-internal nor word-final 152
vowel deletion is obligatory, no such orthographic devices are in place to deal with word-internal vowel deletion, suggesting that word-final deletion is likely more obvious to Kawaiisu speakers.
Aside from the speakers themselves, it seems to be the case that word-final vowel deletion is particularly salient to researchers as well, with Zigmond et al. (1990) writing the following:
“Kawaiisu short vowels in word-final position are often dropped (perhaps more
commonly in rapid speech). Words in which final vowel deletion has occurred are heard
with final stress. This deletion seems to be most frequent at the end of an utterance,
especially in verbs, and is never heard in citation forms.”
(Zigmond et al. 1990:13)
Later, they continue as follows:
“This final vowel loss seems to be a correlate of the consistent devoicing or deletion of
final vowels in other Southern Numic languages… Kawaiisu is unusual within the
subfamily, however, in having mainly fully voiced final vowels and almost no cases of
medial vowel loss.”
(Zigmond et al. 1990:146).
There are a few issues with these statements from Zigmond et al. (1990). First, contrary to the assertion above by Zigmond et al. (1990), it has been discussed in Section 3.4.1 above that word-medial vowel loss in Kawaiisu is fairly common and largely follows patterns that are similar to other systems of vowel reduction in many other languages. As noted, 149 such instances were identified in reasonably careful speech. If conversational or otherwise less careful 153
speech had been analyzed, the rate of medial vowel loss would likely have been even higher, so it is simply not accurate to state that Kawaiisu is lacking in medial vowel loss.
Additionally, Zigmond et al. (1990:13) claim that word-final vowel deletion is “never heard in citation forms.” Quantifying any such assertion with “never’ is generally unsafe, and it is quite untrue to claim that word-final vowel deletion “never” occurs in citation forms. There are numerous examples of word-final vowel deletion in Klein’s field recordings (Klein 1958, 1981-
1983), including in wordlist elicitation sessions where citation forms were often elicited. Perhaps
Zigmond et al. (1990) worked with a different set of data in which no such instances were noted, but, as will be discussed, Klein’s (1958, 1981-1983) recordings show that word-final vowel deletion can occur in citation forms.
Finally, Zigmond et al. (1990) mention that word-final vowel deletion occurs most frequently in verbs. This contrasts with word-internal vowel loss, in which phonological environment (and not morphosyntactic or lexical category) is often relevant in predicting the likelihood of vowel deletion. In the Klein data, 246 instances of word-final vowel deletion were noted. Words where the final vowel was immediately preceded by an obstruent were most commonly affected. In the spectrograms, this is usually accompanied by a very obvious stop burst followed by silence (as seen in Figure 13 above). In words where the penultimate phoneme was a fricative or affricate, high intensity fricative noise was apparent. Due to the variant audio quality across the tapes in the archives, I have not attempted to analyze the acoustic quality of these stop bursts or frication noise associated with these instances of word-final vowel deletion.
The vowel most commonly affected by word-final deletion is /ü/, accounting for approximately 59% of all cases. This was followed by /i/ (32%), /a/ (8%), and /u/ (1%). 154
Ogasawara (2007) suggests that high vowels in a language may be more susceptible to deletion because, cross-linguistically, high vowels tend to have a shorter duration than non-high vowels.
Figure 1 in this chapter indicates that this tends to be true for unstressed high vowels in Kawaiisu as well. Additionally, these vowels are very common in word-final position in Kawaiisu, being found at the end of a majority of Kawaiisu absolutive and diminutive suffixes which often end
Kawaiisu words. The /a/ phoneme often occurs at the end of verbal clitics which have a transitivizing effect, and so may be more common in Kawaiisu discourse than in simple elicitation of words (I direct the reader back to Chapter 1 of this dissertation for a very brief discussion of Kawaiisu grammar, or to Zigmond et al. (1990) for a more thorough reading).
Finally, one claim made by Zigmond et al. (1990) is that word-final vowel deletion tends to affect verbs most frequently. An examination of the Klein recordings (Klein 1958, 1981-1983) casts some doubt on this claim. Slightly over half (54%) of the instances of word-final vowel deletion that were noted in these recordings actually occur in nouns. Only 43% of recorded cases of word-final vowel deletion come from verbs. 11 The remaining 3% of instances come from polarity particles like /juˈwaatü/ ‘no’ and temporal adverbs. Thus, if word-final vowel deletion favors one lexical category over another (which is unlikely, in general), it may well be the case that nouns, rather than verbs, are the favored category for this type of deletion.
However, despite the fact that the Klein data seems to disconfirm the idea that verbs are more commonly affected by word-final vowel deletion, this claim by Zigmond et al. (1990) cannot be dismissed entirely, for a few reasons. First, Zigmond et al. (1990) worked with
11 Some of these verbs are elicited by Klein as adjectives, but function as stative verbs in Kawaiiisu. This is reinforced by Zigmond et al. (1990:60), who note that such “adjectives are indistinguishable in morphology and syntax from ordinary verbs.” 155
speakers who were not represented in Klein’s (1958, 1981-1983) recordings, and since this is not an obligatory phonological process, the ways in which word-final vowel deletion manifests in
Kawaiisu may vary between idiolects. Second, the fact that there were more instances of word- final deletion in nouns than verbs in Klein’s elicitation sessions may reflect a ratio of elicited nouns to elicited verbs which differs substantially from forms elicited by Zigmond (although, since recordings of Zigmond’s elicitations are not available, a direct comparison of his elicitations to Klein’s cannot be made).
Additionally, recall that only data from Klein’s elicitation sessions were considered.
Zigmond et al. (1990) are quite clear in noting that word-final vowel deletion happens most frequently in rapid speech, and utterance-finally. In a wordlist elicitation scenario, the researcher asks for a word, and then the speaker produces the word, at which point the utterance ends, making the end of each word an utterance-final position where word-final vowel deletion is likely to occur. This creates a speech situation which may be somewhat artificial, as words are not usually produced in isolation, but rather tend to comprise longer utterances. Considering such longer utterances, Zigmond et al. (1990:15) also note that Kawaiisu “may be viewed as being dominantly SOV in type.” Free word order is possible in Kawaiisu, but if the language is, as
Zigmond et al. (1990) claim, dominantly SOV, then it is a reasonable assumption that many sentences are verb-final. Assuming the end of an utterance often aligns with the end of a sentence, and the end of a sentence often coincides with a verb, and word-final vowel deletion is most often heard in utterance-final position, then the claim that word-final vowel deletion tends to most frequently affect verbs should not be surprising. In order to test this claim in a more thorough and accurate way, it would be necessary to examine more natural forms of Kawaiisu 156
speech: conversations, stories, longer sentences, etc. However, such an analysis would be prohibitively difficult without the aid of a native Kawaiisu speaker.
Finally, it is important to note that this process of word-final vowel deletion has correlates in other Uto-Aztecan and Numic languages, as Zigmond et al. (1990) mention. Zepeda
(1983) writes that certain vowels in Tohono O’odham (another Uto-Aztecan language, distantly related to Kawaiisu) are shortened, and in fact almost whispered by many O’odham speakers
(implying voicelessness). These vowels, which usually occur word-finally in Tohono O’odham, are represented orthographically with a breve diacritic over the vowel, as in <‘uwĭ> /ʔuwi̥ /
‘woman’ 12 . A similar situation is found in the closely related Akimel O’otham language, where vowels are often devoiced word-finally (Hughes 2016). Voiceless vowels also occur in Southern
Ute, identified by Oberly (2008). In Southern Ute orthography, Oberly (2008) notes that voiceless vowels are written as underlined when the vowel quality is apparent, but when vowel quality is indeterminate, a capital
One key difference between voiceless vowels in Kawaiisu and those found in other Uto-
Aztecan languages is that word-final vowel devoicing in Kawaiisu seems to be optional. In the
Piman languages Tohono O’odham and Akimel O’otham, voiceless vowels seem to often be lexically specified. Oberly (2008) notes that the phonological properties of voiceless vowels in
Southern Ute may be unclear: these vowels may be underlyingly voiceless, or the devoicing which occurs may be due to phonological and morphological processes. Even though the phonological/morphological properties which condition vowel devoicing in Southern Ute may be somewhat unclear, it does not appear to be optional to the extent that it is optional in Kawaiisu.
12 This is the orthographic convention used in Zepeda (1983). This is in contrast to Saxton et al. (1983), in which there is no orthographic distinction between voiced and voiceless vowels. 157
Later, the phonetic realizations of Southern Ute voiceless vowels are investigated by Oberly and
Kharlamov (2015). They find that Southern Ute speakers employ at least three different strategies for vowel devoicing: reduction in intensity and duration of the vowel with lengthening and increased voicelessness of a preceding consonant, partial devoicing with visible formant structure and a weakly present voicing bar, or full devoicing with limited formant structure and no apparent voicing bar. As Southern Ute and Kawaiisu are closely related languages of the
Southern Numic branch, it is quite possible that the phonetic realization of voiceless vowels in
Southern Ute is similar to the vowel deletion process in Kawaiisu discussed in this chapter.
However, the audio in the Klein recordings (Klein 1958, 1981-1983) is not well-suited for this type of analysis, and as such, I leave this question open for future investigation.
3.5. Discussion and Conclusion In this chapter, I have examined several facets of the unstressed vowel system of
Kawaiisu. Vowels which are phonemically long were shown to have a greater duration than vowels which are phonemically short. Tests also revealed that some vowels simply have a shorter inherent duration than others, regardless of phonemic length. For example, the high vowels (with the exception of long /üː/) are shown to be shorter on average than non-high vowels. The greater duration for non-high vowels may be due to the lowering of the tongue body and/or jaw which is often part of the articulation of non-high vowels. The findings regarding the differences in duration between high and non-high vowels are expected. In the articulation of non-high vowels, the jaw tends to be more open than it is during the articulation of high vowels, and this opening of the jaw may contribute to the greater duration seen for non-high vowels
(Lindblom 1967, Crystal and House 1988a, Ohala and Ohala 1992, Ogasawara 2007, etc. ). 158
In addition to tests relating to vowel duration, each speaker's long and short vowel pairs were compared to test for differences in quality. That is, for a given speaker, the formants of /i/ were compared to those of /iː/, the formants of /a/ were compared to those of /aː/, etc. With the exception of RB, each speaker had at least some vowel pairs where statistically significant differences were found. Of those three speakers, several consistent patterns were found. For example, for FC, LG, and AP, the differences in F1 between short /a/ and long /aː/ were all statistically significant, with long /aː/ being lower in the vowel space. In every case where significant differences are found, the long vowels were closer to the peripheral edges of each speaker's vowel space. In fact, a glance at the vowel plots presented in this chapter shows that even where statistically significant differences were not found between the formants of some vowel pairs, the long member of the pair often leans toward the periphery regardless. In these situations, it is often the case that significant differences were not found due to the small token count for some of the vowels (for some cases, the token count was too small for statistical analysis at all). In any case, the general trend, even for unstressed vowels, is for the long member of a vowel pair to fall closer to the edge of the vowel space, while the short member of the pair tends to occupy a more central position in the vowel space. This is expected, as noted by
Lindblom (1963), who notes that as vowel duration increases, speakers have more time to reach a more peripheral acoustic target where the effects of consonantal co-articulation are less prevalent. Essentially, the vowel space shrinks in phonemically short vowels relative for the space for phonemically long vowels. As duration increases, the effect of consonantal co- articulation decreases, and speakers are able to reach more peripheral acoustic targets. 159
Tests were also done to examine the ways a vowel's position in the vowel space moves throughout its articulation. Each speaker showed statistically significant variation for at least some of their vowel phonemes. Some general tendencies were found for all speakers. Where statistically significant variation occurs, it tends to be the case that the unstressed vowels rise in the vowel space over time. This is true for most of the vowels for most of the speakers. Some vowels tend to lower slightly before rising. This is true of /a/ for all four speakers, of /o/ for AP and LG, and of /u/ for LG. Although differences in F1 were more common, some statistically significant differences across timepoints were found for F2. In these instances, it is usually the case that F2 lowers, with vowels moving further back in the vowel space. However, some vowels move forward for some speakers. These include /ee/ for speakers LG and RB, and /i/ for speaker
LG. Again, the forward movement of /eː/ may be associated with the very common verbal suffix
/-kweː/.
This chapter has also examined phenomena of unstressed vowel deletion, in both word- medial and word-final position. Contrary to Zigmond et al. (1990), it has been shown that medial vowel loss does occur in Kawaiisu. Word-medially, vowels (especially high vowels) tend to be lost when adjacent to a voiceless consonant. As discussed, this is not at all unusual in the languages of the world, with a wide range of languages undergoing very similar processes, some of which are discussed in Beckman (1994).
Finally, I discuss word-final vowel deletion in Kawaiisu. Vowels in this position are almost always unstressed, and are often deleted. Word-final vowel deletion in Kawaiisu is both frequent enough and perceptually obvious enough that speakers often write word-final vowels in parentheses to indicate that they may not always be present. Final vowel deletion in Kawaiisu 160
has also been described by Zigmond et al. (1990), writing that this process most frequently occurs in utterance-final position, in rapid speech, and on verbs. However, in the data examined here, there were more instances of word-final deletion on nouns than on verbs. The wordlist elicitation sessions examined in this chapter are not necessarily representative of natural speech, resulting in a situation where the end of each word is the end of an utterance, and if it is the case that final vowels are deleted most often in utterance-final contexts, elicitation of individual words may artificially inflate the frequency of word-final vowel deletion. Further regarding the
Zigmond et al. (1990) assertion that word-final vowel deletion disproportionately affects verbs, I argue that if such a tendency exists in conversational Kawaiisu, it may be due to a coincidence relating to the order of syntactic constituents rather than anything inherent about verbs themselves. Zigmond et al. (1990) note that Kawaiisu is predominately an SOV language. If it is the case that sentences in Kawaiisu often end in verbs, as they do in SOV languages, and if it is true, as Zigmond et al. (1990) note, that word-final deletion occurs most often utterance-finally, then we should expect word-final deletion to affect verbs most often, as verbs in an SOV language are likely to occur in utterance-final position.
From a broader perspective, the results presented in this section raise some issues.
Phonemically long vowels tend to have a larger vowel space than phonemically short vowels in
Kawaiisu, which is in sync with other discussions of stress, vowel duration, and the overall size of the vowel space (for instance, see Tiffany 1959, Lindblom 1963). The data here were presented in isolation, elicited as words in a list. It seems possible that, in conversational contexts, or even in the context of complete sentences, the vowel space would continue to become smaller as rate of speech increases. Additionally, it has been suggested in this chapter 161
that, counter to Zigmond et al. (1990), it may not be the case that vowel deletion is triggered specifically by certain lexical categories. Rather, vowel deletion is likely to occur utterance- finally, and their claim that this phenomenon occurs most frequently in verbs is coincidental, due to a tendency for verbs to occur utterance-finally in an SOV language. With the aid of a native
Kawaiisu speaker to help in the analysis of more spontaneous speech (in conversational or sentential contexts), such questions would be more easily addressed, but given the level of endangerment, such an undertaking may be difficult. 162
CHAPTER 4: ON THE ACOUSTIC CORRELATES OF STRESS IN KAWAIISU
4.1 Introduction In this chapter, I compare the stressed vowel data discussed in Chapter 2 to the unstressed vowel data discussed in Chapter 3, to examine the acoustic correlates of word-level stress in
Kawaiisu, focusing on differences between stressed and unstressed vowels to gain insight into the ways Kawaiisu speakers signal that a particular syllable carries stress. As discussed in previous chapters, stress in Kawaiisu typically falls on the penultimate syllable of a word.
However, there are some instances in which stress can occur on the final syllable. Zigmond et al.
(1990) note that words ending in long vowels have word-final stress. Aside from this, the process of word-final vowel deletion can sometimes give the appearance of word-final stress. If a word is affected by the process of word-final vowel deletion (discussed in Chapter 3 of this dissertation), then the syllable which is normally penultimate becomes, in effect, the word-final syllable.
Because the penultimate syllable usually carries stress, and the process of word-final vowel deletion causes the underlyingly penultimate syllable to surface word-finally, stress can seemingly appear on a word-final syllable (at least at the surface level). An example is given below in (1), repeated from Chapter 1. Figure 13 in Chapter 3 shows an example of a spectrogram containing word-final deletion in the word /seːˈɣidü/
163
(1) Stress and Final Vowel Deletion
(a). [.wi.ná.pi]
(b). [.wi.náp.]
Two possible pronunciations of this word are presented in (1a, b). Periods are used to separate syllables in this example, and stress is marked with an acute accent. In (1a), a word-final
[i] is present. In this example, the final syllable which surfaces is [.pi.], which does not carry stress. In (1b), the final [i] is deleted, resulting in the stressed syllable [.náp.] surfacing in word- final position. Without word-final vowel deletion, the stressed syllable surfaces in penultimate position as in (1a), but when the process does occur, stress appears in the word-final surface syllable. As discussed in previous chapters, none of the stressed vowel data examined in this dissertation come from syllables which are underlyingly word-final, although some may come from words which have been affected by word-final vowel deletion.
On linguistic stress, Hayward (2000:273) writes that stress is both “straightforward to analyse impressionistically” and “difficult to define in purely phonetic terms.” She notes that stress can be defined colloquially in terms of prominence: stressed syllables are said to be somehow more prominent than unstressed ones. This idea of the prominence of stressed syllables has a basis in physical reality; the articulatory mechanisms behind the production of the stressed segments are well-studied. As Ladefoged (2006:23) points out, “variations in stress are caused by an increase in the activity of the respiratory muscles (so that a greater amount of air is pushed out of the lungs).” This increase in muscular energy has acoustic consequences. Ladefoged
(2006), Hayward (2000), and many others point out that the increase in the volume of air released from the lungs is associated with greater intensity (perceived as loudness), and greater 164
duration. Additionally, the laryngeal muscles are often involved in the production of stressed syllables, resulting in variations of fundamental frequency (perceived as pitch) as well. Thus, stressed syllables (and the vowels which comprise their nuclei) are often louder, longer, and have higher pitch than unstressed ones. A given language can use any or all of these to mark prominent syllables, if it has a stress distinction.
In addition to our knowledge of the articulatory mechanisms of stress (discussed for instance in Ladefoged 1963, Lieberman 1967, Kent and Netsell 1971, Ohala 1977, Ohala et al.
1979, de Jong 1991, de Jong 1995, Ladefoged 2006, etc. ), and the corresponding acoustic effects, the perception of stress has also been studied. As mentioned, stressed syllables tend to have greater intensity, greater duration, and higher pitch than unstressed syllables. Numerous studies exist discussing the way native speakers use each of these acoustic cues to determine whether or not a syllable is stressed. Fry (1955) examines duration and intensity as perceptual cues to English stress, finding that duration is a more robust cue than intensity. Fry (1958) expands on his earlier work by considering fundamental frequency in addition to duration and intensity, finding that fundamental frequency outweighs duration as a perceptual cue to stress.
Lieberman (1960) reports similar results, noting that fundamental frequency appears to be a very important perceptual cue to lexical stress (although, counter to Fry (1955, 1958), he finds that intensity may play a greater role than duration). Later, Morton and Jassem (1965) also find pitch, or fundamental frequency, to be the primary correlate of stress, while also identifying some methodological concerns with Lieberman’s (1960) research on stress which may help explain why his results regarding the importance of intensity were so out of sync with some of his contemporaries (i.e. Fry 1955, 1958). In more recent decades, research has extended into L2 165
stress perception (Kawagoe 2002, Nguyen and Ingram 2005, and Wang 2008, for example), a useful avenue of research in a revitalization context, as many speakers of indigenous languages like Kawaiisu may be heritage learners or L2 speakers themselves.
The studies mentioned above focus specifically on speakers of English. Berinstein (1979) finds that the phonological typology of a language plays a role in the ways in which speakers of a language use duration, intensity, and pitch to signal the presence or absence of linguistic stress.
Berinstein (1979:2) hypothesizes, “in a tone language, variations of F 0 of the type used in making the contrast will be the least important cue for determining stressed syllables,” and “in a language with a phonemic length contrast, duration will be the least important.” That is, in a tone language, variations in pitch already signal a phonemic contrast and it is not likely that such variation is also used to signal differences in stress. Similarly, in a language with phonemically long and short vowels (like Kawaiisu), it is predicted that duration should be relatively unimportant as an acoustic correlate of stress, because such variations in duration are used to signal the phonemic contrast between long vowels and short vowels.
Berinstein (1979) studies stress in several languages, finding support for this hypothesis.
For example, in a comparison of K’ekchi and Cakchiquel, two Mayan languages, she finds that duration is relatively unimportant as an acoustic cue for stress in K’ekchi, but maintains its importance as an acoustic correlate of stress in Cakchiquel. The important difference between these two languages is that vowel length is contrastive in K’ekchi, but not in Cakchiquel, and thus, speakers of K’ekchi do not use variation in vowel duration as a robust acoustic cue to stress, because it is used to signal the vowel length contrast instead. 166
It may also be possible for languages to have no stress system. One possible example of such a language is Yucatec Maya. In Yucatec Maya, both length and tone are involved in a phonemic contrast: vowels can be either short, or long. Long vowels can have a high tone or a low tone. In her study of prominence and suprasegmentals in Yucatec Maya, Kidder (2013) finds no evidence for the existence of a stress system in this language, based on both speaker intuitions and acoustic data. This finding supports Berinstein’s (1979) hypothesis on stress and suprasegmentals. If a language has variations in both pitch and duration which interact with each other phonemically, and these are cross-linguistically common cues to stress, then it is not likely that the language will use these cues to signal stress or prominence. Kidder (2013) argues that there may still be a stress system present in Yucatec Maya, but that it is not signaled through the cross-linguistically unmarked acoustic cues of variations in vowel duration or pitch. Navajo, which has phonemic tone and vowel length, has also been argued to lack linguistic stress
(McDonough 2003, Kidder 2008).
Kawaiisu is a non-tonal language. Under Berinstein’s (1979) hypothesis, it can be predicted that variation in pitch can be used as a robust cue to stress in this language. However, as a language with phonemically contrastive vowel length, duration is less likely to serve as a reliable acoustic cue for stress in Kawaiisu.
Oberly (2008) contains an investigation of the acoustic correlates of stress in Southern
Ute. Like Kawaiisu, Southern Ute is a member of the Southern Numic sub-family of the Uto-
Aztecan language family. The two languages are thus closely related and are typologically similar in many ways, with respect to their phonological systems. Both languages are non-tonal, and both have a contrast in vowel length. Thus, according to Berinstein’s (1979) hypothesis, 167
pitch should be a high indicator of stress in Southern Ute, while duration should be less important. Oberly’s (2008) work supports this hypothesis, finding that there were no significant differences in duration between stressed and unstressed vowels in Southern Ute. In fact, unstressed vowels often had a greater duration than stressed ones (this was especially true for long vowels, but no statistically significant differences were reported).
Oberly (2008) finds that pitch in Southern Ute is a more reliable indicator of stress than duration. For most of the speakers she observes, she finds that stressed syllables tend to have a higher pitch than unstressed syllables. One speaker from this study consistently showed higher pitch in unstressed long vowels and lower pitch in unstressed short vowels, but this was the only speaker who broke the pattern (for this speaker, unstressed long vowels were notably greater in duration than stressed long vowels). While statistical significance was not quite attained, Oberly
(2008:152) finds a “near significant effect of stress on subjects’ pitch,” reporting a p-value of p =
0.052. She notes that her analysis may have suffered due to a very small number of speakers, and that an increase in speakers would likely have resulted in statistical significance. Additionally, with only a small number of speakers, a single speaker with aberrant data has the potential to significantly affect a statistical analysis. Regardless, Oberly (2008) is clear in noting that the relationship between stress and pitch is much stronger in Southern Ute than the relationship between stress and duration, with stressed vowels tending to have a higher pitch. Given that vowel length is contrastive in Southern Ute, these findings are in line with Berinstein’s (1979) hypothesis.
In the remainder of this chapter, I focus on duration and fundamental frequency as acoustic correlates of stress in Kawaiisu. I also compare the formant structures of stressed and 168
unstressed vowels, testing for differences in vowel quality. Intensity is not analyzed here, for two primary reasons. First, the quality of the audio in the available recordings (Klein 1958, 1981-
1983) is highly variable, and there is no reliably consistent reference intensity level to compare to, making any results of an intensity analysis uninterpretable. Second, one goal of this description of word stress in Kawaiisu is to determine whether Kawaiisu matches Berinstein’s
(1979) predictions (as Southern Ute does, based on Oberly’s (2008) work). Namely, if a language makes use of a suprasegmental feature as part of a phonemic contrast, then that suprasegmental feature should not function as a reliable indicator of stress, because it is already used to signify a different type of contrast. Under Berinstein’s (1979) hypothesis, the way a language uses these suprasegmentals in phonemic contrasts determines the way stress is realized in the language. Pitch and stress are used by many languages phonemically: some languages have phonemic tone, some languages have phonemic vowel length, and some languages have both. No language has been reported to use variations in intensity or loudness as part of a system of phonemic contrasts. Thus, the question of how the stress system of a language interacts with phonemic contrasts in intensity is not a valid one.
4.2. Methods The data in this chapter were seen previously in Chapters 2 and 3 of this dissertation. For detailed explanations of measurement criteria and the criteria by which tokens were chosen for inclusion in these analyses, see the methodology sections of those chapters (2.2, 3.2). In this chapter, stressed syllables are those which occur in penultimate position, as discussed in Section 169
4.1 above. Vowels in other positions (with the exception of word-final long vowels) are unstressed.
Briefly, Praat (Boersma and Weenink 2017) software was used to extract formant values
(and fundamental frequency) from each vowel token which was selected for analysis, at three separate points in time (20%, 50%, and 80% of the duration of each vowel token). The duration of each vowel token was also measured. In this chapter, the duration of stressed vowels is compared to the duration of unstressed vowels to determine the influence stress has, if any, on vowel duration in Kawaiisu. Surrounding consonantal context is one additional factor that may play a role in the duration of vowels, however the data are not well enough distributed to control for surrounding consonantal context, and as such, I have not attempted to measure the relationship between vowel duration and surrounding consonantal context.
Differences in formant structure (vowel quality) and fundamental frequency (pitch) between stressed and unstressed vowels are also tested. In the tests comparing vowel quality of long and short vowel pairs seen in Chapters 2 and 3 of this dissertation, I averaged across timepoints so that, instead of each token having three separate measurements for each formant, only the averages were considered. That is, for some token /a/, rather than feeding three separate pieces of information into the analysis (one measurement for the 20% timepoint, one for 50%, and one for 80%), I took the average of those three measurements, considering only those averages in the analyses. I take the same approach in this chapter, in tests comparing fundamental frequency and formant structure between stressed and unstressed vowels. This approach reduces the influence of outliers and any measurement errors. 170
4.3. Results 4.3.1 Duration
Below, Figure 1 shows the difference in duration between stressed and unstressed vowels.
FIGURE 1: Comparison of duration between stressed and unstressed vowels, averaged across all vowels for all speakers. Phonemically short vowels appear in the left panel, phonemically long vowels in the right. Unstressed vowels are represented by blue bars, and stressed vowels by green.
In the figure above, it is apparent that stressed vowels are generally longer in duration than unstressed vowels, but that the effect of stress is far smaller than the effect of phonemic 171
vowel length. One exception here is short /e/, which is shown to have a longer duration on average when it is unstressed. While speaker RB did not produce any stressed tokens of /e/ that fit the criteria for inclusion in the analysis (see Chapter 2 for a discussion of this issue), this pattern was consistent for each of the other speakers. For all speakers who did produce analyzable tokens of both stressed and unstressed short /e/, the unstressed member of the pair had a greater duration on average.
While the remaining vowels do show a pattern in which the stressed members of the pairs have greater durations on average than the unstressed members of each pair, these differences tend to be rather small. The average difference between a stressed short vowel and an unstressed short vowel is only about 10ms. The average difference between a stressed long vowel and an unstressed long vowel is somewhat higher, at about 25ms. In Figure 1, large differences in duration are seen between stressed and unstressed /oː/, and stressed and unstressed /uː/. It may be the case that these differences are artificially high due to the small number of tokens for these vowels (seen in Chapters 2 and 3, there were only 9 tokens of stressed /oː/, 12 of stressed /uː/, 29 of unstressed /oː/, and 31 of unstressed /uː/), and these averages could be smaller if more individual tokens had been included.
To compare stressed and unstressed vowel durations, a repeated measures ANOVA was conducted. In this analysis, the data were averaged such that, for example, the measurement given for a speaker’s short unstressed /i/ duration is the average duration of all tokens of short unstressed /i/ from that speaker. The dependent variable was duration in milliseconds. There were three factors: stress (2 levels: stressed and unstressed), phonemic length (2 levels: long and short), and vowel type (with 5 levels: one for each vowel type except /e/, /eː/, because one 172
speaker did not produce any analyzable tokens of short stressed /e/). There was a significant main effect of stress on duration ( F(1, 3) = 137.466, p = 0.001). There were also significant main effects of both vowel type ( F(4, 12) = 19.598, p < 0.001) and phonemic length ( F(1, 3) =
110.569, p = 0.002). Significant interactions were found between stress and vowel type ( F(4, 12)
= 4.615, p = 0.017), stress and phonemic length ( F(1, 3) = 15.348, p = 0.030), and vowel type and phonemic length ( F(4, 12) = 11.008, p = 0.001).
To examine the interaction between stress and phonemic length, a pair of separate repeated measures ANOVAs were conducted: one examined only the long vowels, and the other only examined the short vowels. In these tests, there were two levels: stress (with two levels, stressed and unstressed) and vowel type (5 levels, one for each vowel quality excluding /e/). For the short vowels, there was a significant main effect of stress ( F(1, 3) = 24.558, p = 0.016). For the long vowels, there was also a significant main effect of stress ( F(1, 3) = 79.360, p = 0.003).
The differences in F-ratio and significance level indicate that, while stress effects the duration of both long and short vowels, the difference in duration between stress levels is greater for long vowels.
To examine the interaction between stress and vowel type, additional tests were conducted, this time including the /e/ vowel type. In all, six of these tests were done (one for each vowel type). These tests had two factors each: stress and phonemic length, each with two levels (stressed vs. unstressed, and long vs. short, respectively). Of these six tests, significant differences in stress were found only for the /o/ type ( F(1,3) = 172.737, p < 0.001) and the /u/ type ( F(1,3) = 29.120, p = 0.012). Altogether, the results of the tests on duration and stress indicate that a general pattern does exist wherein the average duration of stressed vowels is 173
greater than that of unstressed vowels, but this tendency is stronger for long vowels. However, testing each vowel type separately reveals that stress only plays a significant role in duration for the /o/ and /u/ types, which is consistent with Figure 1, which shows relatively large differences in duration between stressed and unstressed /oː/, and stressed and unstressed /uː/. These large differences may account in large part for the significant main effect of stress that was found in the initial repeated measures ANOVA that was conducted in this section.
4.3.2 Pitch
Below, in Figure 2, a general pattern is shown in which stressed vowels, for each speaker, have a higher pitch than unstressed vowels. The male speaker, FC, has the lowest pitch averages, while the three female speakers have higher pitch. In the figure above, it can also be seen that variations in pitch are greater between short vowels than between long vowels. For each speaker, the pitch difference between stressed and unstressed long vowels is smaller than the pitch difference between stressed and unstressed short vowels, which can be seen in the figure. 174
FIGURE 2: Comparison of pitch between stressed and unstressed vowels. Pitch values are in Hz. Short vowels are shown in the left panel, long vowels in the right panel. Unstressed vowels are represented by blue bars, while stressed vowels are represented by green bars. Speakers are identified by initials across the bottom of the chart.
To test the effect of stress on pitch in Kawaiisu, I conducted a repeated measures
ANOVA with three factors: stress (2 levels: stressed and unstressed), phonemic length (2 levels: long and short), and vowel type (5 levels, one for each vowel excluding /e/, due to RB’s lack of short unstressed /e/ tokens). Data is averaged over items (words) so that each speaker has a single value for each condition before analysis. That is, this is effectively a by-speakers
ANOVA. There was a significant main effect of stress ( F(1, 3) = 97.975, p = 0.002), of length 175
(F(1, 3) = 38.773, p = 0.008), and of vowel type ( F(4, 12) = 4.291, p = 0.022). There was also a significant interaction of stress and phonemic length ( F(1, 3) = 66.960, p = 0.004). To examine this interaction further, a pair of additional ANOVAs was conducted. One test compared short vowels only, and the other compared long vowels only. Stress was found to significantly affect the pitch of both short vowels ( F(1,3) = 104.432, p = 0.002) and long vowels ( F(1,3) = 56.449, p
= 0.005). Overall, this is consistent with Figure 2, in which it can be seen that, for each speaker, pitch is higher in stressed vowels, although the difference in pitch between stressed and unstressed vowels is greater in vowels which are phonemically short.
Due to speaker RB’s lack of short stressed /e/ tokens, the /e/ vowel type was tested separately from the analyses above. RB’s data were excluded from the comparisons involving the /e/ vowel type. This was a repeated measures ANOVA with stress and phonemic length as factors, each with two levels. There was a significant main effect of stress on pitch ( F(1, 2) =
132.143, p = 0.007). The effect of phonemic length was not significant ( F < 1), and neither was the two-way interaction.
Finally, additional tests were done to compare the relationship between stress and vowel type for the remaining vowels aside from /e/. These tests took the same form as the test detailed above for the /e/ type (repeated measures ANOVAs with stress and phonemic length as factors).
All speakers’ data were included in these remaining tests. Stressed vowels were found to have significantly higher pitch for the /a/ vowel type (F(1, 3) = 348.366, p < 0.001), the /i/ vowel type
(F(1, 3) = 69.534, p = 0.004), the /o/ vowel type ( F(1, 3) = 44.878, p = 0.007), and the /u/ vowel type ( F(1, 3) = 19.125, p = 0.022). There was a near-significant effect of stress on pitch for the
/ü/ vowel type ( F(1, 3) = 9.240, p = 0.056). 176
4.3.3 Vowel Quality
In this section, I compare each speaker’s stressed and unstressed vowel space, testing the effects of stress on vowel quality. The vowel plots in this section are similar in appearance and form to those seen in previous chapters (for instance, triangles still represent long vowels and squares still represent short vowels). The one major difference between these vowel plots and previous ones is that here, I do not consider separate timepoints. The data used in this section have been averaged across timepoints such that each vowel token has only a single measurement for each formant (rather than three measurements for each formant). Since separate timepoints are not considered in these analyses or included in the vowel plots, differences in color are meant to represent different levels of stress, rather than measurements taken at different timepoints.
Here, blue represents stressed vowels, and red represents unstressed vowels. First, an F1 x F2 vowel plot for FC is shown in Figure 3, comparing his stressed and unstressed vowel spaces. 177
FIGURE 3: F1 x F2 plot comparing the stressed and unstressed vowels of speaker FC. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
Above, in Figure 3, it can be seen that most of this speaker’s stressed and unstressed vowels have similar averages. There are a few noticeable differences. Unstressed /ü/ appears somewhat higher and further forward than stressed /ü/, which is much closer to the center of the vowel space. Differences also occur between stressed and unstressed /uː/, and stressed and unstressed /oː/. In both of these cases, the unstressed versions have lower F2 averages. In some instances, the unstressed member appears to lie closer to the edge of the vowel space than the stressed member. This can be seen, for instance, for /uː/, /oː/, /eː/, and /üː/. Next, I show the F1 x
F3 vowel plot for speaker FC. 178
FIGURE 4: F1 x F3 plot comparing the stressed and unstressed vowels of speaker FC. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
In Figure 4, differences in F3 between stressed and unstressed vowels can be seen.
Relatively large F3 differences are seen between stressed and unstressed /iː/, /eː/ and /oː/. Next, I test stressed and unstressed vowel pairs for differences in quality. To investigate the relationship between vowel quality, I conducted some by-items ANOVAs on speaker FC’s vowel data. As in previous chapters, I test each speaker separately, as methods of vowel space normalization may not be reliable given the small number of speakers. Each stressed and unstressed vowel phoneme are tested separately, so, for example, stressed /a/ is compared to unstressed /a/, stressed /aː/ is compared to unstressed /aː/, etc. The dependent variables in these tests were F1, F2, and F3, all averaged over timepoint such that each token had a single value for each formant, which was the 179
mean of the formant measurements at 20%, 50%, and 80% of the vowel’s duration. In each of the subsequent tests in this chapter, vowels with fewer than 5 tokens are not included in the statistical analyses. F-ratios and significance are shown below in Table 1. For vowels where statistically significant differences are found in either F1 or F2, a separate column indicates whether the stressed or unstressed member of the pair is more peripheral.
TABLE 1: F-ratios and significance levels comparing the effect of stress on vowel quality (F1, F2, and F3) for each vowel phoneme for speaker FC. Significance is marked as indicated, and appears in bold. Vowel Degrees of F1 F2 F3 Which is more Phoneme freedom peripheral? /a/ 1, 147 1.099 4.599* 7.601** Unstressed (lower) /aː/ 1, 23 F<1 F<1 F<1 N/A /e/ Excluded from analysis due to paucity of data /eː/ 1, 32 12.457** 28.250*** 18.959*** Unstressed (higher, more front) /i/ 1, 84 1.270 2.284 10.775** N/A /iː/ Excluded from analysis due to paucity of data /o/ 1, 59 F<1 F<1 10.233** N/A /oː/ Excluded from analysis due to paucity of data /u/ 1, 56 11.914** F<1 F<1 Unstressed (higher) /uː/ 1, 8 F<1 27.853** F<1 Unstressed (more back) /ü/ 1, 46 3.405 8.985** 1.399 Unstressed (higher) /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
A few significant differences are found in Table 1 above. Stressed and unstressed /a/ differ significantly from each other in both F2 and F3, with unstressed /a/ having a more centralized average, and a higher F3 average than stressed /a/. Long /eː/ differs significantly 180
across all three formants, with unstressed /eː/ being higher and further forward in the vowel space, in addition to having a higher F3. Unstressed short /u/ appears higher in the vowel space than stressed short /u/, long /uː/ differs more in the front-back dimension, with unstressed /uː/ being further back than stressed /uː/. Finally, the difference between stressed and unstressed short /ü/ is shown to be significant for F2, but not for any other formants, with unstressed /ü/ being more centralized than the stressed version. The rightmost column explains the direction of effect for vowels where significant differences were found. In several cases, unstressed vowels were more peripheral than stressed vowels. Next, the stressed and unstressed vowels of speaker
AP are compared. Her vowels, and the vowels of all other speakers in this study will be presented and analyzed as FC’s were above. 181
FIGURE 5: F1 x F2 plot comparing the stressed and unstressed vowels of speaker AP. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
Above, I show the stressed and unstressed vowel space for speaker AP. Many of the stressed/unstressed vowel pairs have averages which are close together in the figure, although there are a few which are reasonably well-separated (such as the /aː/, /oː/, and /ü/ pairs). Like
FC’s vowels, some of speaker AP’s unstressed vowels fall closer to the periphery than their stressed counterparts. Below, Figure 6 shows an F1 x F3 plot of speaker AP’s vowels. 182
FIGURE 6: F1 x F3 plot comparing the stressed and unstressed vowels of speaker AP. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
A few differences in F3 between stressed and unstressed vowel pairs can be seen here.
For example, the /iː/ pair is quite well separated in F3, with the unstressed member of the pair having an F3 average approximating 3200Hz, while stressed /iː/ has a noticeably lower F3 average of around 2950Hz. Similarly, unstressed /uː/ seems to have a very high F3, around
3250Hz, while stressed /uː/ is much lower, averaging closer to 2900Hz. The results of the statistical analyses for speaker AP’s vowels are shown below in Table 2. 183
TABLE 2: F-ratios and significance levels comparing the effect of stress on vowel quality (F1, F2, and F3) for each vowel phoneme for speaker AP. Significance is marked as indicated, and appears in bold. Vowel Degrees of F1 F2 F3 Which is more Phoneme freedom peripheral? /a/ 1, 186 F<1 1.080 4.381* N/A /aː/ 1, 48 7.823** 1.184 F<1 Unstressed (lower) /e/ Excluded from analysis due to paucity of data /eː/ 1, 98 10.116** 1.263 1.026 Unstressed (higher, more front) /i/ 1, 181 3.550 2.773 3.393 N/A /iː/ 1, 69 2.615 1.034 3.709 N/A /o/ 1, 78 F<1 2.333 F<1 N/A /oː/ Excluded from analysis due to paucity of data /u/ 1, 49 F<1 F<1 8.035** N/A /uː/ Excluded from analysis due to paucity of data /ü/ 1, 83 F<1 4.254* 1.149 Unstressed (more back) /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
The results in Table 2 are consistent with the information shown in Figures 5 and 6 above. For instance, Figure 5 shows that the /a/ stressed-unstressed pair is similar in F1 and F2, but Figure 6, which includes F3, shows a difference in F3 of approximately 100Hz. Likewise, it can be seen that the /aː/ pair is similar in both F2 and F3, but has a larger separation of F1, with unstressed /aː/ closer to the lower edge of the vowel space. A similar situation obtains for the /ee/ pair, with unstressed /eː/ closer to the periphery than stressed /eː/. Finally, Table 2 indicates that the /ü/ pair was significantly different in F2, but not F1 or F3, and the figures above show that this pair is very similar in F1 and F3, but has a much larger difference in F2. Regarding this vowel, the unstressed member of the pair was further back (and slightly higher) than the stressed 184
pair. Like FC, speaker AP has several unstressed vowels which are closer to the periphery than stressed vowels. Next, the stressed and unstressed vowels of speaker LG are examined.
FIGURE 7: F1 x F2 plot comparing the stressed and unstressed vowels of speaker LG. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
The stressed and unstressed pairs of speaker LG seem to be a bit more separated than they are for other speakers. In the figure above, it can be seen that most stressed vowels are not particularly close to their unstressed counterparts. Below, Figure 8 shows the same speaker’s vowels in an F1 x F3 plot. 185
FIGURE 8: F1 x F3 plot comparing the stressed and unstressed vowels of speaker LG. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
In Figure 8, it can be seen that many of speaker LG’s stressed and unstressed vowel pairs are reasonably separated in terms of F3 as well: for example, stressed and unstressed /iː/ are very well separated, as are stressed and unstressed /aː/, and many others. Generally, it seems to be the case that most of the stressed vowels have an F3 average between 2800-2900Hz, while the unstressed vowels tend to have a slightly lower F3 on average, with most of them falling in the
2700-2800Hz range. The results of statistical tests that were conducted on this speaker’s vowel data are shown below in Table 3.
186
TABLE 3: F-ratios and significance levels comparing the effect of stress on vowel quality (F1, F2, and F3) for each vowel phoneme for speaker LG. Significance is marked as indicated, and appears in bold. Vowel Degrees of F1 F2 F3 Which is more Phoneme freedom peripheral? /a/ 1, 385 12.356*** 8.000** 6.028* Stressed (lower) /aː/ 1, 46 16.874*** 1.453 F<1 Stressed (lower) /e/ Excluded from analysis due to paucity of data /eː/ 1, 39 116.102*** 24.290*** F<1 Unstressed (higher, more front) /i/ 1, 287 38.931*** 32.102*** 4.608* Stressed is more front, unstressed is higher /iː/ 1, 25 6.313* 1.960 4.297* Unstressed (higher) /o/ 1, 164 22.548*** F<1 3.468 Unstressed (higher) /oː/ Excluded from analysis due to paucity of data /u/ 1, 118 32.631*** 8.970** 24.167*** Unstressed (higher, more back) /uː/ Excluded from analysis due to paucity of data /ü/ 1, 163 54.469*** 1.181 4.291* Unstressed (higher) /üː/ 1, 12 2.467 F<1 F<1 N/A * p < .05 ** p < .01 *** p < .001
For speaker LG, every vowel pair that was tested (except for long /üː/) showed a statistically significant difference between levels of stress in at least one formant. Generally, this fits with the vowel plots for LG seen above, where the separations between stressed and unstressed vowel pairs in the vowel space are visually obvious. The only vowel pairs where the stressed members were closer to the periphery were /a/, /aː/, and possibly /i/. Lastly, the stressed and unstressed vowel spaces of speaker RB are considered, with vowel plots and statistical tables shown below. 187
FIGURE 9: F1 x F2 plot comparing the stressed and unstressed vowels of speaker RB. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
Here, it is seen that speaker RB’s /a/ and /aː/ stressed and unstressed vowel pairs are reasonably close together, but the separation is generally greater for the other stressed and unstressed pairs. Especially large distances are seen between the /üː/ vowels, the /uː/ vowels, and the /i/ vowels, for instance. Below, Figure 10 compares the stressed and unstressed vowels of speaker RB along the F1 x F3 dimensions. 188
FIGURE 10: F1 x F3 plot comparing the stressed and unstressed vowels of speaker RB. Squares represent short vowels, triangles represent long vowels. Stressed vowels are shown in blue, and unstressed vowels are shown in red.
In this plot, it can be seen that F3 varies based on stress for some vowel pairs, but not for others. For instance, the /aː/ pair is similar in F3, as is the /o/ pair. Others, such as the /ü/ pair, are much more greatly separated in terms of F3. The results of statistical tests that were conducted on speaker RB’s vowel data are shown below in Table 4. 189
TABLE 4: F-ratios and significance levels comparing the effect of stress on vowel quality (F1, F2, and F3) for each vowel phoneme for speaker RB. Significance is marked as indicated, and appears in bold. Vowel Degrees of F1 F2 F3 Which is more Phoneme freedom peripheral? /a/ 1, 94 F<1 F<1 1.474 N/A /aː/ 1, 41 3.074 F<1 F<1 N/A /e/ Excluded from analysis due to paucity of data /eː/ 1, 36 3.962 F<1 2.105 N/A /i/ 1, 77 19.927*** 2.437 2.242 Stressed is more front; unstressed is higher /iː/ Excluded from analysis due to paucity of data /o/ 1, 43 4.004 F<1 F<1 N/A /oː/ Excluded from analysis due to paucity of data /u/ 1, 26 6.264* F<1 1.499 Unstressed (higher) /uː/ Excluded from analysis due to paucity of data /ü/ 1, 69 3.676 3.169 6.551* N/A /üː/ Excluded from analysis due to paucity of data * p < .05 ** p < .01 *** p < .001
Many of the vowel pairs could not be tested for differences in vowel quality because there were some vowels for which speaker RB did not produce a number of tokens that was sufficient for analysis. This is not the fault of the speaker herself; rather, it was simply the case that some vowels were not well-represented in the words that were elicited in her recordings. Of the vowel pairs where statistical analysis was possible, significant differences in vowel quality were found for the /i/ pair, the /u/ pair, and the /ü/ pair. The /i/ and /u/ pair are shown to differ significantly in terms of F1, or vowel height. Although unstressed /i/ is more centralized along the F2 dimension than stressed /i/, the differences in F2 were not significant. The height difference was significant, with unstressed /i/ being higher in the vowel space (closer to the 190
periphery) than stressed /i/. Unstressed /u/ is also significantly higher in the vowel space than stressed /u/, and thus closer to the periphery. For /ü/, the difference is in F3, and Figure 10 shows that unstressed /ü/ has an F3 average that is approximately 150Hz higher than the F3 average for stressed /ü/.
4.4. Discussion and Conclusion In this chapter, I compared the stressed and unstressed vowels of four Kawaiisu speakers, focusing on three aspects in particular: vowel duration, vowel pitch, and vowel quality. Duration was found to be only somewhat reliable as an acoustic correlate of Kawaiisu stress. Although there is a general pattern for stressed vowels to have a greater duration than unstressed vowels
(this is true for both phonemically long and phonemically short vowels), these differences tend to be relatively minor. Analysis of individual vowel qualities indicated that the only vowel qualities where stress played a significant role in duration were the /o/ and /u/ qualities. That is, stressed
/oː/ and /uː/ are greater in duration than their unstressed counterparts, but the effect of stress on the duration of other vowel qualities is negligible. There is a general tendency for stressed vowels to be greater in duration, but outside of the two vowel qualities mentioned above, these differences tend to be rather small.
Stressed vowels in Kawaiisu were found to have higher pitch than unstressed vowels.
This is true regardless of phonemic length: both long and short vowels were shown to have higher pitch when they carry lexical stress. Additionally, most vowel types were shown to have higher pitch when stressed. The only individual vowel type where this is not certain is the /ü/,
/üː/ type. Statistical analysis revealed that the difference in pitch between stressed and unstressed 191
vowels of this vowel type was very nearly, but not quite, significant. Generally, the /üː/ phoneme
(in both stressed and unstressed conditions) had a low token count. It is possible that if this phoneme had been represented better in the data, and the test had thus had more statistical power, analysis would have revealed a significant difference in pitch across stress level for this vowel type as well. This would fit with the pattern that stressed vowels in Kawaiisu, regardless of phonemic length or vowel type, tend to have a higher pitch than unstressed ones. Considering these facts, along with the facts surrounding pitch and stress in this language, it appears to be the case that Berinstein’s (1979) observations on stress and suprasegmental features hold true for
Kawaiisu. The data show pitch to be a more reliable indicator of Kawaiisu stress than vowel duration. As Berinstein (1979) predicts, a suprasegmental feature that a language uses in a phonemic contrast is not likely to serve as a strong indicator of a contrast in stress. Variations in duration are used in Kawaiisu to signal a phonemic contrast between long and short vowels, but variations in pitch are not used in any phonemic contrast in this language. The Kawaiisu data fit the expected pattern: the suprasegmental feature which is involved in a phonemic contrast is not a strong indicator of stress, but the feature which is not involved in such a contrast is a more reliable acoustic indicator of stress.
I also examine differences in vowel quality and formant structure between stressed and unstressed vowels. While each speaker had several vowel phonemes which differed in formant structure depending on stress, there were no consistent patterns where, for example, one phoneme or another varied the same way across the majority of speakers. However, many of the tests showed some unstressed vowels lying closer to the edge of the vowel space than some 192
stressed vowels. A table summarizing these results, including information about which vowel was more peripheral, and direction of effect, is shown below.
TABLE 5: Summary of differences in vowel quality between stressed and unstressed vowels for each speaker Vowel FC AP LG RB Phoneme /a/ Unstressed is No difference Stressed is No difference lower lower /aː/ No difference Unstressed is Stressed is No difference lower lower /e/ Not tested Not tested Not tested Not tested /eː/ Unstressed is Unstressed is Unstressed is No difference higher and more higher and more higher and more front front front /i/ No difference No difference Unstressed is Unstressed is higher, stressed higher, stressed is is more front more front /iː/ Not tested No difference Unstressed is Not tested higher /o/ No difference No difference Unstressed is No difference higher /oː/ Not tested Not tested Not tested Not tested /u/ Unstressed is No difference Unstressed is Unstressed is higher higher and more higher back /uː/ Unstressed is Not tested Not tested Not tested more back /ü/ Unstressed is Unstressed is Unstressed is No difference higher more back higher /üː/ Not tested Not tested No difference Not tested
In many cases, there were no differences in quality between stressed and unstressed vowel pairs for these speakers. For the 12 vowel phonemes examined here, FC showed no difference in 3 pairs, and both AP and RB showed no difference in 5 pairs. LG only had one pair which did not differ in quality. When differences were found, it was usually the case that the unstressed member was closer to the periphery. This was the case for 5 of FC’s vowel pairs, 3 of 193
AP’s vowel pairs, 5 of LG’s vowel pairs, and 1 of RB’s vowel pairs. The /i/ vowel pair for both
RB and LG was somewhat ambiguous, as the stressed member was more front (with a higher
F2), and the unstressed member was higher in the vowel space (with a lower F1). However, the only speaker which had vowel pairs where the stressed member was unambiguously more peripheral was LG, whose /a/ and /aː/ pairs both had stressed members with averages lower in the vowel space than the unstressed members. Generally, there were no clear patterns across speakers: some vowels did not display a difference in speakers, while other vowels did. In vowels where differences were found, even when the unstressed member was more peripheral, those differences were not always realized the same way: for FC and LG, unstressed /ü/ was more peripheral than stressed /ü/ due to a difference in vowel height, while for AP, unstressed /ü/ was judged more peripheral due to a difference in vowel backness. The /ü/ vowel pair for FC and
LG did not differ in backness, and AP’s /ü/ vowel pair did not differ in height. The only instance where the variation seems to be uniform is in the /eː/ pair. For three of the four speakers, unstressed /eː/ was both higher, and further front in the vowel space. Because there were often no differences in quality between stressed and unstressed vowel pairs, and the differences that do exist do not usually display uniform patterns of variation across speakers (although the unstressed versions are often shown to be more peripheral), I hold that the Kawaiisu stress system is characterized by a lack of vowel reduction. Unstressed vowels in Kawaiisu are not reduced, and can be as peripheral (if not moreso) as the stressed vowels.
In Chapters 2 and 3, it was shown that phonemically long vowels, whether stressed or unstressed, are greater in duration than phonemically short vowels. It was also shown in both of these previous chapters that phonemically long vowels tend to lie closer to the edge of the vowel 194
space than phonemically short vowels. This seems to suggest a relationship between greater duration and greater peripherality. That is, the longer a vowel’s duration, the easier it is for a speaker to reach a more peripheral acoustic target, where the acoustic effects of consonantal co- articulation are limited. This was argued for by Lindblom (1963:1780), who writes, “in spite of efforts on the part of the talker to hit the bull’s-eye articulation, he cannot do so at fast rates, owing to the limitations inherent in the articulatory mechanism… Duration seems to be the main determinant of the reduction.” Vowels which are shorter in duration should lie in a more central position, while vowels which are longer in duration should fall closer to the edge. A similar pattern has been shown in Kawaiisu: among vowels of the same stress level, phonemically long vowels are both greater in duration and more peripheral in the vowel space than phonemically short vowels.
However, many studies have also identified that linguistic stress tends to be associated with greater vowel duration (see for instance Fry 1955, Lieberman 1960, and other works cited in this chapter). Thus, if stressed vowels usually have greater duration, and vowels with greater duration tend to be more peripheral in the vowel space, then it is expected that stressed vowels should lie closer to the edges of the vowel space. Comparing the vowels of American English in isolated, stressed, and unstressed conditions, Tiffany (1959) finds that the vowel space effectively shrinks as levels of stress and duration decrease, with unstressed vowels tending to be the most centralized. Other studies (Gay 1978, Fourakis 1991, for example) show similar results.
In Kawaiisu, on the other hand, it is sometimes the case that unstressed vowels in Kawaiisu are closer to the edges of the vowel space. 195
There are a few possible reasons for this apparent discrepancy. First, it has been documented that unexpected linguistic patterns can frequently arise in situations involving severe states of language endangerment (discussed in several chapters in Dorian 1989). Also related to language endangerment is the fact that only a small handful of speakers were able to be recorded.
It is possible that, with a larger number of speakers and more consistent wordlists specifically tailored to address questions relating to stress and vowel quality, different patterns may have emerged.
On the other hand, it is possible that this pattern has little to do with language endangerment or the limited quantity of data. Above, it was discussed that vowels with shorter durations tend to be more centralized as speakers undershoot a more peripheral acoustic target
(Tiffany 1959, Lindblom 1963). One of the findings of this chapter was that duration plays only a limited role in Kawaiisu stress. The differences in duration between stressed and unstressed vowels tend to be fairly small. If vowel undershoot is tied to duration, and the relationship between duration and stress is tenuous in Kawaiisu, then it should not be surprising to find that
Kawaiisu sometimes fails to fit the expected pattern regarding the relationship between level of stress and peripherality or centralization of vowel quality. Further, as noted by Berinstein (1979), the link between duration and stress is not expected to be as strong in a language with a phonemic vowel length distinction like Kawaiisu. However, the data are not clear enough to suggest a greater peripherality of unstressed vowels in general, because the patterns of which unstressed vowels are more peripheral are not consistent across speakers. Some speakers may show certain unstressed vowels as being more peripheral, while other speakers may have an entirely different set of unstressed vowel phonemes being more peripheral. Further, many of the 196
vowel pairs which were tested showed no statistically significant differences in F1 or F2 at all.
Thus, while it is true that some speakers had some unstressed vowels with formant averages that were closer to the vowel space edges, it is perhaps more accurate to claim that, in terms of quality, unstressed Kawaiisu vowels are characterized by a lack of reduction, rather than greater peripherality. Because vowel duration is not strongly linked to stress in Kawaiisu, then this finding is not unexpected. 197
CHAPTER 5: ON CONSONANTS
5.1 Introduction This chapter presents an analysis of the consonants of Kawaiisu. In particular, I investigate voice onset time (henceforth VOT) of the voiceless stop consonants. In addition, I examine the acoustic characteristics of the velar phoneme labeled as /g/ in Klein (1959, 2002) and Zigmond et al. (1990) (and labeled as /ɣ/ in this dissertation), showing that this phoneme is only rarely articulated as a true occlusive. Other concerns relating to consonant articulations were mentioned in Chapter 1 of this dissertation, and are revisited in greater detail in the present chapter. These include palatalization of coronal fricatives and affricates in environments containing the high front vowel, acoustic characteristics of the glottal stop, acoustic and articulatory variation in rhotic consonants, and the phenomena of pre-aspirated and glottalized sonorants in word-medial position. Many of the sounds examined in this chapter are cross- linguistically rare (pre-glottalized and pre-aspirated nasals, for instance), and have not received much phonetic documentation or description in any language. For those sounds which are more cross-linguistically common (such as rhotic trills and flaps, and palatalized sibilants), documenting them in Kawaiisu tells us more about the possible range of ways languages can make distinctions. In many instances, I show that the consonants described in this chapter are somewhat variable in their possible pronunciations, which may be useful in language revitalization work.
198
5.2 The Velar Obstruent As discussed in Chapter 1 of this dissertation (shown in Tables 1 and 3 of Chapter 1), both Klein (1959, 2002) and Zigmond et al. (1990) note the existence of a Kawaiisu phoneme /g/
(notated as /ɣ/ in the present work). Zigmond et al. (1990) elaborate, claiming that, although they choose to represent the sound with the /g/ symbol, standardly used to represent a voiced velar stop, the sound in question is actually articulated as a voiced velar fricative [ɣ]. Klein (2002) has a similar impression of this sound, noting that it is often a spirant or otherwise weakly articulated. As noted by Ohala and Riordan (1979) and Ohala (1983, 1997), there are aerodynamic reasons for this weakening of /g/ (discussed in greater detail later in this section), and patterns of /g/ becoming a fricative or approximant are cross-linguistically well-attested. In this section, I use acoustic data to present a brief discussion of the various ways speakers may produce this sound.
The /ɣ/ (as well as the labialized /ɣw/) phoneme exhibits some variability in pronunciation. In the relatively careful speech represented in Sheldon Klein’s wordlist elicitation sessions (Klein (1958, 1981-1983), see Chapter 1 of this dissertation for more discussion of this data source), two main variants occur: a stop [g], and a fricative [ɣ]. In some cases, the segment is so weakly articulated that it may become more approximant-like, being produced occasionally as [ɰ]. In this section, I do not attempt to quantify the amount of turbulence in the acoustic signal to distinguish between strongly articulated fricatives, weakly articulated fricatives, and approximant-like productions. In a more controlled setting, such a task would be possible (likely with the finding that degree of constriction is related to speech rate or other temporal factors), but given the limitations and quality of the available data, this is prohibitively difficult. Thus, I 199
divide the productions of /ɣ/, /ɣw/ into stops and fricatives (or non-stops), with the understanding that the non-stop variants can be approximants.
In the Klein wordlist elicitation sessions, a total of 437 articulations of either /ɣ/ or /ɣw/ were noted. One of the key acoustic characteristics of oral stop consonants is the stop release burst, a short, aperiodic, transient, impulse-like burst of sound which occurs as the pressure behind the consonantal closure is released. As noted by Crystal and House (1988b), this burst is not always visible in a spectrogram. They note that release bursts tend to be the most visible with voiced stops, and that velars tend to be “complete” (meaning there is both an obvious closure and release burst) more often than other places of articulation. In the Kawaiisu data, I note 4 possibilities for the articulation of the /ɣ/, /ɣw/ phonemes: stops with visible bursts, stops without visible bursts, fricatives, and approximants.
Of the 437 articulations of these phonemes that were counted, stops with visible bursts comprised the smallest portion. Bursts were only visible on 52 of the 437 /ɣ/, /ɣw/ tokens, roughly 12%. Examples are shown below in Figures 1 and 2.
200
FIGURE 1: Spectrogram and waveform of /tüˈɣahni/
201
FIGURE 2: Spectrogram and waveform of /seːˈɣidü/
In Figures 1 and 2 above, the portions of the spectrograms containing the release bursts for the /ɣ/ phonemes (and for the /t/ in Figure 1) show that, in these utterances, these particular phonemes were articulated clear stops, complete with visible release bursts. There were many instances of the /ɣ/, /ɣw/ phonemes, where a consonantal stop closure was apparent, but bursts were not visible. These account for 96 of the 437 counted tokens, or roughly 22% of the total. In these instances, no frication noise was apparent in the spectrogram, nor was any formant activity visible which would indicate an approximant articulation. An example of a stop with closure, but without burst, is shown below in Figure 3. 202
FIGURE 3: Spectrogram and waveform of /seːˈɣidü/
Articulations with obvious fricative noise were more common, totaling 107 of the 437 counted tokens of /ɣ/, /ɣw/ (approximately 24% of the total). An example of a fricative articulation for /ɣ/ is shown below in Figure 4, with visible fricative noise highlighted. 203
FIGURE 4: Spectrogram and waveform of /paɣiˈkweːdü/
In the remaining articulations, some formant activity was visible during the articulation of the /ɣ/, /ɣw/ phonemes, indicating that these were produced as approximants. These were 183 of the 437 instances of these phonemes, or roughly 42%. One such example is shown below in
Figure 5. 204
FIGURE 5: Spectrogram and waveform of /tuɣuˈkwidü/
While Figure 4 displays visible frication in the spectrogram during /ɣ/, no such fricative turbulence can be seen in the spectrogram for Figure 5. The visibility of formants combined with the lack of frication noise associated with /ɣ/ indicates that the /ɣ/ in Figure 5 is likely an approximant articulation. While the presence of a stop burst is quite apparent in a spectrogram
(as seen in Figures 1 and 2 above), it is sometimes difficult to characterize the differences between the other articulations mentioned here. Phonetic differences are often gradient rather than categorical, and it is not always straightforward to determine how much fricative noise is 205
necessary to call an articulation a fricative, or how much formant activity is necessary to conclude that an articulation is an approximant. There were some instances where frication noise and formant activity were both visible within a single articulation, or instances where formant structure was partially visible during articulations that were mostly more stop-like. The exact numbers given here for different articulations may differ depending on the stringency of the criteria for what counts as a given articulation type. It is possible that the quality of the preceding vowel may play a role in whether the /ɣ/ phoneme is articulated more like a stop, more like a fricative, or more like an approximant, but the data are not sufficient to make a clear generalization. Regardless, the general pattern suggests that the /ɣ/, /ɣw/ phonemes are articulated as stops in roughly 1/3 of instances, while the remaining 2/3 of articulations were more fricative-like or approximant-like. Thus, I hold that Zigmond et al. (1990) are essentially correct in noting that Kawaiisu /ɣ/ is not usually a true stop. However, some variation is apparent such that even though Kawaiisu /ɣ/ is often produced as a fricative or approximant, stop articulations also occur, at least in the careful speech represented in the Klein wordlist elicitations.
Phonological gaps such as this, in which a language is missing voiced velar stops from the phoneme inventory, are well-attested and fairly common. Sherman (1975) notes that when a language has an asymmetrical inventory of stop phonemes where a voiced stop is missing, the absent stop is often velar. There are phonetic explanations for this type of development. For instance, Ohala (1983, 1997), and Ohala & Riordan (1979) explain that voicing can only be maintained when air is flowing over the glottis, causing it to vibrate. When enough air flows over the glottis that air pressure above the glottis reaches the same level as pressure below the glottis, 206
air can no longer flow past the glottis, resulting in cessation of voicing. The above sources note a few primary reasons why it is difficult to maintain voicing in velar stops. First, the distance between the glottis and the velar closure is relatively small. Because of this, it does not take much time for supraglottal air pressure to become equal to subglottal air pressure, resulting in short duration of voicing. In addition to this, there are various means of increasing the amount of time that voicing can be maintained, but this becomes more difficult the further back the consonantal closure is. For example, expanding the vocal tract can increase the amount of time voicing can be maintained. This can be done by lowering the larynx. This can also happen when the surface of the vocal tract between the glottis and the consonantal closure is relatively compliant (such as the cheeks during the articulation of a labial stop). In the case of the velar stop, this area is limited to the pharyngeal wall and soft palate, which are not compliant enough to yield much to air pressure. Because of the difficulties in maintaining voicing for velar stops, languages may lose voiced velar stops from their phoneme inventories altogether. Finally, Ohala
(1997) explains that intervocalic voiced stops have a tendency to become fricatives or approximants (in which supraglottal air pressure does not rise as high). In Kawaiisu, the only position where the /ɣ/ phoneme occurs is intervocalically. Considering the difficulties in maintaining voicing of back-articulated stops such as velars, and the tendency of intervocalic stops to weaken to fricatives or approximants, it is not surprising that Kawaiisu /ɣ/ tends to surface in this way. Thus, the apparent anomaly in the asymmetry of the Kawaiisu stop system can be explained as a natural phenomenon which arises from well-described aerodynamic concerns.
207
5.3 Voice Onset Time Kawaiisu has been claimed (Zigmond et al. 1991) to have a voicing distinction in its stop series. In Kawaiisu, those are the voiceless /p, t, k, kw/, and the voiced /b, d/ (the above section explains the reason for not treating /ɣ, ɣw/ as stops). This appears to be a binary categorical distinction: stops are either voiced or voiceless. Thus, Kawaiisu has only a two-way VOT distinction in its stop series. The concept of voice onset time is discussed in much more detail in
Lisker and Abramson (1964).
While voiceless stops can occur both word-initially and word-medially in Kawaiisu, the voiced stops are distributionally restricted, occurring in intervocalic, word-medial position only.
Kawaiisu voiced stops are fully voiced. Vocal fold vibration is maintained throughout the articulation of a preceding vowel, during the articulation of a voiced stop, and following the release burst into a following vowel. Voicing does not cease during this sequence. An example is shown below in Figure 6. 208
FIGURE 6: Spectrogram and waveform of /joziˈküdü/
Above, Figure 6 shows a spectrogram and waveform of /joziˈküdü/
voiced stop in Kawaiisu. Voicing is maintained throughout the closure, and voiced stops only occur in word-medial, intervocalic position.
Ladefoged (2003) mentions a few possible criteria for measuring VOT. In this chapter, I measure from the beginning of the release burst to the onset of periodicity. Since voiced stops in
Kawaiisu tend to be fully voiced, I do not measure their VOT. If VOT of voiced stops were measured, VOT would be equivalent to the duration of the consonantal closure in almost all instances involving voiced stops. Duration of these closures would likely depend on speech rate, speaker, and other factors, and in order for these results to be interpretable or meaningful, the duration of voiced stop closures would need to be compared to duration of other articulations.
While voiced stops in Kawaiisu are universally fully voiced, VOT of voiceless stops in
Kawaiisu is more variable. It has been shown that, in many languages, VOT of voiceless stops varies depending on factors such as consonantal place of articulation, following vowel quality, position in the word, and whether or not the consonant occurs in a stressed or unstressed syllable
(see for instance Cho and Ladefoged 1999, Ladefoged 2003, Bijankhan and Nourbaksh 2009,
Berry and Moyle 2011). For instance, it has been noted that VOT tends to be greater on stops occurring word-initially, or at the beginning of a stressed syllable (Lisker and Abramson 1967,
Ladefoged 2003). In the remainder of this section, I examine the four voiceless stops in
Kawaiisu, /p, t, k, kw/, to determine which factors play important roles in the duration of VOT in
Kawaiisu. Below, Table 1 shows token counts for each stop, for each speaker, in each of the relevant conditions (initial vs. medial word position, stressed vs. unstressed syllables, and following vowel length). 210
TABLE 1: Number of stop consonant tokens available for VOT analysis. Table is organized by speaker on the vertical axis, and by phoneme on the horizontal axis. Individual cells show the total token counts for each stop phoneme for each speaker, including token counts for specific conditions. Speaker Consonant phoneme /p/ /t/ /k/ /kw/ AP 147 total 137 total 117 total 29 total 86 initial 98 initial 49 initial 3 initial 61 medial 39 medial 68 medial 26 medial 42 stressed 28 stressed 45 stressed 19 stressed 105 unstressed 109 unstressed 72 unstressed 10 unstressed 132 before short V 108 before short V 88 before short V 10 before short V 15 before long V 29 before long V 29 before long V 19 before long V FC 95 total 116 total 66 total 23 total 53 initial 100 initial 43 initial 3 initial 42 medial 16 medial 23 medial 20 medial 23 stressed 33 stressed 34 stressed 17 stressed 72 unstressed 83 unstressed 32 unstressed 6 unstressed 89 before short V 88 before short V 51 before short V 12 before short V 6 before long V 28 before long V 15 before long V 11 before long V LG 237 total 277 total 148 total 49 total 136 initial 186 initial 62 initial 7 initial 101 medial 91 medial 86 medial 42 medial 56 stressed 39 stressed 59 stressed 38 stressed 181 unstressed 238 unstressed 89 unstressed 11 unstressed 220 before short V 247 before short V 135 before short V 28 before short V 17 before long V 30 before long V 13 before long V 21 before long V RB 46 total 83 total 66 total 28 total 24 initial 53 initial 34 initial 2 initial 22 medial 30 medial 32 medial 26 medial 5 stressed 13 stressed 15 stressed 16 stressed 41 unstressed 70 unstressed 51 unstressed 12 unstressed 41 before short V 69 before short V 52 before short V 8 before short V 5 before long V 14 before long V 14 before long V 20 before long V Totals 525 total 613 total 397 total 129 total 299 initial 437 initial 188 initial 15 initial 147 medial 137 medial 209 medial 114 medial 126 stressed 113 stressed 153 stressed 90 stressed 399 unstressed 500 unstressed 244 unstressed 39 unstressed 482 before short V 512 before short V 326 before short V 58 before short V 43 before long V 101 before long V 71 before long V 71 before long V
211
In the table above, it can be seen that /p/ and /t/ are fairly common stops in Kawaiisu, with the velar and labiovelar stops appearing much less frequently. The /p/ and /t/ stops occur most often in initial position, while /k/ and /kw/ occur more often in medial position. As seen in previous chapters, short vowels occur more frequently than long vowels, so it is expected that stops preceding short vowels should have a higher token count on average, but /kw/ tends to precede long vowels more often, due to the high frequency /-kweː/ suffix, which contains a long vowel.
Below, Figure 7 shows a series of graphs displaying VOT averages for each of the voiceless stop phonemes of Kawaiisu, in the same conditions listed above in Table 1: word position (initial vs. medial), level of stress (stressed vs. unstressed), and following vowel length. 212
FIGURE 7: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme, word position, stress level, and following vowel length. The data have been averaged across all four speakers.
Above, Figure 7 shows the average VOT duration (in milliseconds) for each of the voiceless stop phonemes of Kawaiisu across a variety of conditions. To examine these data further, a series of three repeated measures ANOVAs were conducted. Each had two factors. The first factor, consonant identity, was constant in all three tests, with one level for each voiceless stop phoneme (/p, t, k, kw/, for a total of four levels in this factor). The remaining factors were word position (two levels: initial vs. medial position), stress level (two levels: stress vs. unstressed), and following vowel length (two levels: following long vowel vs. following short 213
vowel). The reason for breaking these up into several tests was that not all speakers produced tokens in certain environments, creating difficulties in certain comparisons. For example, speaker
RB produced zero tokens of /kw/ in initial position followed by a short vowel, and FC produced zero tokens of /kw/ in initial position followed by a long vowel. In the remainder of this section,
I present and discuss the results of these tests, including additional graphs showing VOT for each consonant, collapsed over the factors that are not being tested. The first test examines consonant identity and word position. Below, Figure 8 presents the average VOT for each voiceless stop phoneme varying by word position.
FIGURE 8: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme and word position. The data have been averaged across all four speakers.
214
In the first test (consonant identity x word position), a significant main effect of consonant identity was found ( F(3, 9) = 16.063, p = 0.001), with back-articulated consonants
(/k/, /kw/) having a longer VOT on average. There was no effect of word position ( F(1, 3) =
3.124, p = 0.175), nor was there a significant interaction between these two factors ( F(3, 9) =
2.827, p = 0.099). Next, I consider consonant identity and stress level.
FIGURE 9: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme and stress level. The data have been averaged across all four speakers.
In this test (consonant identity x stress level), a significant main effect of consonant identity was found ( F(3, 9) = 14.591, p = 0.001), again with back-articulated consonants tending to have a longer VOT on average than more front-articulated consonants. There was no effect of stress level ( F < 1). The interaction between these two factors was also not significant ( F(3 ,9) =
2.696, p = 0.109). The final test in this series examines consonant identity and phonemic length of the immediately following vowel. 215
FIGURE 10: Mean VOT duration (in ms) for each voiceless stop phoneme of Kawaiisu. Factors shown are consonant phoneme and following phonemic vowel length. The data have been averaged across all four speakers.
In this final test (consonant identity x following vowel length), there was again a significant main effect of consonant (F(3,9) = 15.208, p = 0.001), with the back-articulated consonants /k/ and /kw/ having a longer VOT on average than the front-articulated consonants
/p/ and /t/ . The effect of vowel length was also significant ( F(1,3) = 157.079, p = 0.001): VOT duration is longer when the following vowel is long, but shorter when the following vowel is short. There was no significant interaction ( F(1,3) = 1.153, p = 0.380).
Overall, VOT tends to be longer for /k/ and /kw/, but smaller in /p/ and /t/. This is consistent with the findings of Fischer-Jørgensen (1954), Peterson and Lehiste (1960), and Cho
& Ladefoged (1999), which hold that VOT is expected to be greater for consonants with back 216
articulations, like the velars. Cho & Ladefoged (1999) discuss possible reasons for this tendency, noting that the reason posterior articulations are associated with longer VOT than anterior articulations depends on factors such as extension of contact between the articulators, and the speed at which the articulators move.
Word position also has an effect on VOT in many languages, with word-initial stops tending to have greater VOT, as mentioned in Keating et al. (2003). Based on visual inspection of previous figures, this is only true in Kawaiisu for certain conditions. For /k/, VOT is greater in initial position, but for /p/ and /t/, the effect of word position, if present at all, is minimal. The difference in VOT for /p/ and /t/ in these positions is minute. The effect of word position on
VOT of /kw/ is unclear, due to a lack of /kw/ tokens preceding short vowels in initial position, but based on the data that are available, there may be some effect such that VOT is greater for
/kw/ in initial position. This is certainly true for /kw/ tokens which precede long vowels, in any case. However, statistical analysis did not reveal a significant effect of word position on VOT in
Kawaiisu.
In many languages, stress also plays a role in VOT duration, with stops beginning stressed syllables tending to have greater VOT duration, as discussed in Lisker and Abramson
(1967). However, this is not consistent in Kawaiisu. Based on visual inspection of the figures above, VOT is greater in stressed syllables for initial /k/, medial syllables containing /kw/, and syllables containing /t/, but there are several conditions where stress seems to have a minimal effect. For instance, stress does not have a noticeable effect in syllables containing initial /p/.
There was no significant effect of stress on VOT in Kawaiisu. 217
Finally, it appears to be the case that the phonemic length of the following vowel does have an effect on VOT. In the majority of cases shown in the figures above, VOT is greater when the following vowel is phonemically long. This is true for both /k/ and /t/, regardless of other factors such as stress level or word position. This is also true for many cases involving /p/
(except in the medial stressed condition). The situation with /kw/, again, is less clear. In medial stressed position, /kw/ tends to have a higher VOT on average when it is followed by a phonemically long vowel, but in unstressed position, VOT is higher when it is followed by a phonemically short vowel. There were no tokens available of /kw/ in stressed word-initial position that were followed by a phonemically short vowel, but word-initial stressed /kw/ in the environment of a phonemically long vowel has a comparatively high VOT, approaching 60ms.
This is the second highest VOT average shown in Figure 7, so it is likely that, if tokens of stressed word-initial /kw/ in a short vowel environment were available, their VOT would be somewhat shorter. In addition, the statistical analyses presented above revealed a significant effect of vowel length on VOT. In Kawaiisu, VOT is longer when a stop is immediately followed by a long vowel.
In summary, Kawaiisu maintains a phonemic voicing distinction between voiced and voiceless stops. This distinction is only relevant in word-medial position, because word-medial position is the only place where both voiced and voiceless stops can occur (only voiceless stops may occur word-initially). Word-medially, voiced stops tend to be fully voiced, with vocal fold vibration maintained throughout their articulation. Regarding word position, there was no significant difference in VOT of voiceless stops between those in initial position and those in medial position. While it is common for VOT to be numerically but not significantly higher in 218
initial position, VOT is not necessary to distinguish between voiced and voiceless stops in this position, because voiced stops do not occur word-initially in Kawaiisu. Stress was not found to have a significant effect on VOT duration. Following vowel length was significant, with stops followed by long vowels having longer VOT.
5.4 Palatalization One phonological phenomenon mentioned in Chapter 1 of this dissertation involves palatalization of coronal fricatives and affricates in environments containing front vowels.
Essentially, /s/ becomes [ʃ], /z/ becomes [ʒ], and /ts/ becomes [tʃ]. The description of this process contained in Zigmond et al. (1990) is not immediately clear from their prose alone, as they write simply that this process occurs “in the environment of front vowels.” From the text, it is not explicit whether this happens before a front vowel, or after a front vowel, or between two front vowels. Examining the examples they provide, however, makes this somewhat clearer. For example, they compare the form [hanaˈsübi]
[ʃ], as it immediately follows the front vowel /i/. Thus, it seems that when Zigmond et al.
(1990:2) refer to this process of palatalization occurring in “the environment of front vowels,” they mean specifically that it happens when the relevant phoneme occurs immediately following a front vowel. In this section, I examine this phenomenon more closely, using acoustic evidence to provide support for this claim that the coronal fricatives and affricates are palatalized when following a front vowel. 219
Jongman et al. (2000), in a study of English fricatives, find that the spectral peaks of fricatives decrease in frequency as the oral constriction moves further back in the mouth. In their study, they note that alveolars have an average spectral peak of 6839Hz, while palato-alveolars have an average spectral peak of 3820Hz. Ladefoged (2003) notes similar patterns between [s] and [ʃ], with [s] having a higher concentration of spectral energy just above 6000Hz, while [ʃ] tends to have more energy in the lower frequencies, between 3000-4000Hz. Both Jongman et al.
(2000) and Ladefoged (2003) discuss numerous techniques of analyzing the acoustics of fricatives to glean information about consonantal place of articulation, including F2 locus equations and analyses of spectral moments, but the available Kawaiisu data are not well-suited for techniques such as these due to concerns like variability in audio quality, recording equipment and setting, and presence of tape hiss and other ambient noise. As Cho et al.
(2002:211) write, “some acoustic characteristics [for fricatives] can be qualitatively observed in the spectrograms.”
As it is attested that the acoustic energy present in fricatives decreases in frequency the further back the consonantal closure is (Jongman et al. 2000, Ladefoged 2003), it is expected that in Kawaiisu, palatalized variants such as [ʃ] should have higher concentrations of acoustic energy at lower frequencies than their non-palatalized counterparts. Thus, following Zigmond et al.
(1990), it can be expected that tokens of the coronal fricatives and affricates which immediately follow a front vowel should have more energy in the lower frequency range than those which follow non-front vowels. In the remainder of this section, I examine spectrograms of words containing the /s/, /z/, and /ts/ phonemes in different vocalic contexts, illustrating that 220
palatalization does often occur when these phonemes immediately follow front vowels, especially /i/.
First, I examine spectrograms where the relevant fricatives and affricates are surrounded on both sides by back vowels, to provide a reference for comparison to those contiguous to front vowels. According to Zigmond et al. (1990), palatalization should not occur in this environment.
Below, a spectrogram of /paˈtsaːtü/
FIGURE 11: Waveform and spectrogram of /paˈtsaːtü/
221
In the spectrogram above in Figure 11, the /ts/ affricate occurs between two back vowels,
/a/ on either side. High amplitude frication noise begins at approximately 4500Hz. As Ladefoged
(2003) notes, palato-alveolar fricatives are often associated with higher concentrations of energy between 3000Hz and 4000Hz. As the fricative noise shown in this spectrogram is above the range Ladefoged (2003) gives for palato-alveolars, I argue that in this spectrogram, palatalization does not occur. Another example is shown below in Figure 12.
FIGURE 12: Waveform and spectrogram of /patsaˈriːdü/
As in Figure 11, the affricate shown in Figure 12 occurs between two back vowels.
During the fricative portion of the affricate, high amplitude frication noise starts at approximately 4000Hz, only slightly less than the frequency seen in Figure 11. Again, this is 222
slightly above the range which Ladefoged (2003) notes is typical for palatalization. Below,
Figure 13 shows an example where the affricate occurs immediately following a back vowel, but immediately preceding a front vowel.
FIGURE 13: Waveform and spectrogram of /noɣotseˈʔedü/
As in Figures 11 and 12, the highest amplitude frication noise during the fricative portion of the /ts/ affricate is above the frequency range which Ladefoged (2003) suggests for palatalization. One key difference between this example and previous examples is that Figure 13 shows the affricate followed by a front vowel. Zigmond et al. (1990) notes that palatalization is triggered in environments containing front vowels, but suggest through their examples, as mentioned above, that the relevant environment is the preceding, rather than the following 223
vowel. In Figure 13, the front vowel follows the affricate, and palatalization does not occur. In the examples to follow, the relevant consonant occurs immediately following front vowels
(either /i/ or /e/), and palatalization does occur.
FIGURE 14: Waveform and spectrogram of /aˈnisi/
In Figure 14, the frication noise starts at a comparatively lower frequency than seen in
Figures 11-13. According to Zigmond et al. (1990), this is a typical environment where palatalization should occur, as the fricative is immediately following the front vowel /i/. The lower frequency frication noise, combined with the fact that /s/ appears to be in the appropriate environment, suggests that the fricative seen in Figure 14 is palatalized. Below, Figure 15 shows a similar example involving the affricate /ts/. 224
FIGURE 15: Waveform and spectrogram of /wiˈhitsi/
As in Figure 14, the relevant consonant in Figure 15 occurs between two front vowels, /i/ on either side. High amplitude frication noise begins at around 2500Hz, lower than the non- palatalized examples seen in Figures 11-13. The high concentration of frication noise at lower frequencies suggest that this is an example of palatalization, and it occurs in precisely the environment described by Zigmond et al. (1990). Figure 16 shows an example where palatalization occurs following /e/, suggesting that any front vowel (not just /i/), can trigger this process. 225
FIGURE 16: Waveform and spectrogram of /weːsaˈɣidü/
In Figure 16, the concentration of acoustic energy for the fricative begins at approximately 2500Hz, which is similar to the examples in Figures 14 and 15. Here, palatalization can be argued to occur, due to the lower frequency frication noise. In Figures 14 and 15, the relevant consonant was surrounded on both sides by a front vowel, but here, /s/ is followed by a back vowel /a/. However, due to the preceding front vowel /eː/, palatalization still occurs.
In total, 539 tokens of either /s/ or /ts/ were recorded, and a variety of environments were represented: these phonemes can occur word-initially (before a front vowel or a back vowel), between two vowels of any backness (back-back, front-front, front-back, or back-front), or 226
between a front vowel and a consonant. 13 Some charts showing token count for each environment and distribution of acoustic energy are shown below in Figures 17-23.
FIGURE 17: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ preceded by a front vowel, and followed by a consonant (in the /-vistü/ suffix).
Above, Figure 17 represents the relevant phoneme occurring in a very specific morphophonemic environment. This is the /s/ phoneme in the /-vistü/ ‘a lot’ suffix. According to
Zigmond et al. (1990), palatalization occurs when /s/ or /ts/ appears in adjacent proximity to a front vowel. Here, /s/ is immediately preceded by the front vowel /i/. All noted tokens of the relevant phoneme except for one show high amplitude acoustic energy beginning in the lower frequency ranges, with most beginning in the 2000-2500Hz range, illustrating that /s/ in this particular environment is likely to undergo palatalization. Next, I examine instances of the /s/,
/ts/ phonemes occurring word-medially between two front vowels in Figure 18.
13 This particular environment (/s/ between a vowel and a consonant) only occurs in a single, specific morpheme: the suffix /-vistü/ ‘a lot’. Generally, underlying consonant sequences are rare in Kawaiisu. This suffix is mentioned again in Section 8 of this chapter in a discussion of Kawaiisu syllable structure. 227
FIGURE 18: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ in the environment between two front vowels.
Again, as noted by Zigmond et al. (1990), palatalization of the /s/, /ts/ phonemes occurs when adjacent to front vowels, in words such as /ʔaˈnisi/
FIGURE 19: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ in the environment between a front vowel (on the left) and a back vowel (on the right).
In Figure 19 above, the /s/, /ts/ phonemes are all still following front vowels, but they precede back vowels. As in previous figures, the majority of tokens in this environment have frication noise beginning at a low frequency range, indicating that the coronal sibilants in
Kawaiisu are receptive to palatalization in this environment. Interestingly, there were a fair number of tokens with frication noise beginning at frequency ranges above 3000Hz in this environment. These instances correspond mostly to /s/ phonemes in loanwords from Spanish where the target phoneme was /s/, such as /sarˈve:sa/
‘table’. Because speakers were attempting to reproduce Spanish /s/ in these loanword examples, these particular tokens show frication noise which begins in slightly higher frequency ranges than those which represent native Kawaiisu words in the same environment. This indicates that if the /s/, /ts/ phonemes are in native Kawaiisu words in this environment, they are likely to undergo palatalization, but if they are in loanwords where the phoneme is not palatalized in the 229
source language, then palatalization is not as likely to occur. Next, I consider these phonemes occurring in the environment between two back vowels, as in /paˈtsaːtü/
/paˈtsariːdü/
FIGURE 20: Token count and distribution of acoustic energy for word-medial tokens of /s/, /ts/ in the environment between two back vowels.
Here, a shift is apparent. In Figures 17-19, the majority of tokens of /s/, /ts/ had frication noise beginning in the lower frequency ranges, but when these phonemes occur between two back vowels, the starting frequency range for frication noise tends to be a little higher, with high amplitude energy for most tokens starting above 3000Hz. This indicates that coronal sibilants in
Kawaiisu are less likely to be as heavily palatalized in this environment. Although the majority of tokens have starting energy for frication noise in the ranges above 3000Hz, quite a few tokens show starting energy in lower frequency ranges as well. The majority of tokens associated with lower frequency ranges in this chart come from a single speaker (FC), whose sibilants tend to have lower frequency frication noise than sibilants produced by other speakers. Although this speaker is the only male in this study, and the only speaker from the 1950s, there is not enough 230
evidence to conclude that this difference in palatalization is due to generational or gender-related factors. This trend for this speaker is consistent across all environments. If this speaker’s data were not considered, the chart above would look different, with only 2 tokens in the range below
2000Hz, 4 tokens in the 2000-2500Hz range, and 11 tokens in the 2500-3000Hz range (ranges starting above 3000Hz would not be affected), with an average starting frequency of approximately 3400Hz. Below, I examine /s/, /ts/ between back vowels (on the left) and front vowels (on the right), such as /noɣotseˈedü/
FIGURE 21: Token count and distribution of acoustic energy for word-initial tokens of /s/, /ts/ immediately after a back vowel and immediately before a front vowel.
Here, the general trend is that /s/, /ts/ tokens in this environment tend to lie in the 3000-
3500Hz range. Again, the data are somewhat skewed towards lower frequency ranges because speaker FC tended to palatalize sounds heavily in this environment, while others did not.
Removing this speaker’s data would result in a single peak in the chart above in the 3000-
3500Hz range, with only 2 tokens in the range below 2000Hz (~2% of the total as opposed to
~7%), and 10 tokens in the 2000-2500Hz range (~12% of the total as opposed to ~23%). Higher 231
ranges would not be affected. Generally, the distribution of acoustic energy illustrated in Figure
21 seems similar to the distribution of energy seen in Figure 20, where /s/, /ts/ occurs between two back vowels. Namely, the starting frequencies of frication noise are somewhat higher in these environments (compared to environments where these phonemes follow front vowels only), indicating that the relevant phonemes are not likely to be as heavily palatalized when they follow back vowels. In the following two charts, I consider instances of /s/, /ts/, which occur word-initially, as in /seːˈɣidü/
FIGURE 22: Token count and distribution of acoustic energy for word-initial tokens of /s/, /ts/ followed immediately by front vowels.
In this chart, some variation is apparent. There does not seem to be an obvious trend towards higher or lower starting frequencies in this particular environment. Of the categories in this chart, the 2000-2500Hz range is the most well-represented, but the difference in token count between this category and others is slight. Below, I consider word-initial /s/ and /ts/ phonemes followed by a back vowel, such as /ˈsuːju/
232
FIGURE 23: Token count and distribution of acoustic energy for word-initial tokens of /s/, /ts/ followed immediately by back vowels.
The figure above illustrates that the majority of /s/, /ts/ tokens occurring word-initially and followed by back vowels show high amplitude acoustic energy in the 3000-4000Hz range.
There were 94 noted tokens of /s/, /ts/ in this environment, and of this total, 63 of them had high amplitude acoustic energy beginning at or above 3000Hz. The average starting point for acoustic energy of coronal fricatives and affricates in this environment was approximately 3100Hz.
However, as noted previously, FC consistently produced /s/, /ts/ tokens (in this environment and all others) with acoustic energy starting at much lower frequencies than all other speakers. If this speaker’s data are removed as outliers, then starting frequencies below 3000Hz represent only about 25% of tokens (as opposed to about 33% of tokens, if this speaker’s data are considered).
Generally, however, the much smaller number of tokens with high amplitude acoustic energy beginning in the lower frequency ranges in the chart above indicates that very few tokens in this environment are likely to be palatalized. 233
The charts in Figures 17-23 show that heavier palatalization of sibilants is more likely to occur in Kawaiisu when these consonant phonemes follow front vowels. Tokens of /s/ and /ts/ following front vowels have frication noise starting at lower frequency ranges when they are following front vowels than when they follow back vowels. Starting frequencies for frication noise are somewhat higher when the preceding sound is a back vowel, indicating that sibilants following back vowels tend to not be associated with palatalization.
In this section, I have expanded on the claim made by Zigmond et al. (1990) that coronal fricatives and affricates are palatalized in environments containing front vowels. This is a claim that benefits from the acoustic evidence presented here, as the prose description of this process in
Zigmond et al. (1990) is not entirely explicit regarding the phonological environment that palatalization requires. Considering several of their examples however, it seems the case that they intend to make the claim that palatalization only occurs in Kawaiisu when a coronal fricative or affricate is preceded by a front vowel. In Figure 13, a spectrogram containing the form /noɣotseˈʔedü/
The lack of palatalization is apparent due to the higher frequencies at which the frication noise is concentrated. Though the /ts/ affricate is adjacent to a front vowel in this example, a non- palatalized allophone surfaces because the front vowel follows, rather than precedes, the affricate. This is in contrast with the example shown in Figure 16, which shows a spectrogram containing the form /weːsaˈɣidü/
environment for palatalization. The acoustic evidence presented in this section supports these claims.
5.5. The Rhotic Consonant According to Zigmond et al. (1990:5), Kawaiisu has one rhotic consonant /r/, which they describe as a “flap.” No further elaboration is given. In this section, I examine acoustic evidence to show that some variation exists in the articulation of this consonant. In fact, it may be the case that the rhotic in Kawaiisu is more commonly realized as a trill, rather than a flap. Acoustically, rhotic consonants are often associated with a decrease in F3. As Ladefoged (2003:149) writes,
“variations in the frequency of F3 indicate the degree of r-coloring: the lower the F3, the greater the degree of rhoticity.” In a language like English, where /r/, in many dialects, has an approximant articulation, a falling F3 may be the most obvious cue to rhoticity. However, flaps and trills, with a greater degree of constriction than a typical approximant, have extra acoustic cues beyond a low F3.
From an articulatory perspective, a flap is similar to a brief stop. The tongue briefly comes into contact with an upper surface of the vocal tract. Typically contact is brief enough that there is no significant air pressure built up behind the closure, and there is no release burst. Trills occur when there is a series of rapid taps where one articulator vibrates against another. During a flap articulation, the tongue strikes the upper surface of the vocal tract only once. Johnson
(2008), in a study of Spanish trills, notes the precise number of occlusions during a trill articulation can vary, but that three occlusions is average (see also Quilis (1981)). Additionally,
Ladefoged (2003) shows spectrograms of Finnish, which has a phonemic contrast between a 235
short trill and a long trill. In the short trill, two separate contacts are counted, while four contacts are counted in the long trill.
In the remainder of this section, I examine spectrograms of Kawaiisu words containing the rhotic /r/, noting the decrease in F3 which is typical of rhotic consonants, and also providing evidence for variation between a flap articulation (with only one contact of the articulators), and a trill articulation (with several).
Figure 24 shows a token of /woʔoˈravi/
FIGURE 24: Waveform and spectrogram of /woʔoˈravi/
In Figure 24 above, the drop in F3 is visible and has been traced in red. This drop in F3 is typical in rhotic contexts (although it has been noted that flap consonants such as this are often associated with a decrease in F4 as well, as discussed by Warner & Tucker (2017)). The /r/ is realized as a flap here: a single brief tap of the tongue to the upper surface of the vocal tract, likely in the dental or alveolar region. This is shown with a red arrow in the spectrogram. In the 237
waveform during this portion, it can be seen that there is a decrease in amplitude associated with this segment, relative to the surrounding vowel sounds. Additionally, the spectrogram is somewhat lighter during this articulation, also indicating a lower amplitude. The /r/ occupies a shorter duration than the surrounding segments, which is characteristic of a flap consonant. The boundary between the /w/ and the following /o/ was placed during a brief silence (as indicated by the lower amplitude in the waveform and the lighter region in the spectrogram), likely due to the glottal stop present in this word.
While the rhotic shown in the spectrogram above appears to be a single tap, it is also common for rhotic consonants in Kawaiisu to have a trilled articulation. Below, in Figure 25, I present another spectrogram from speaker AP, illustrating a rhotic with two contacts between the articulators. 238
FIGURE 25: Waveform and spectrogram of /hoˈrodü/
In the example above, the dip in F3 is noticeable on the /o/ preceding the /r/, indicating rhoticity. Two arrows are shown which indicate the spots in the sound signal where the tongue likely contacts the upper surface of the vocal tract during the articulation of the /r/ consonant.
These arrows coincide with a decrease in amplitude, indicating a brief closure where air does not flow as freely from the vocal tract. As it is the case that a flap involves only a single contact, but trills involve more than one, the figure above is an example of a trilled articulation of Kawaiisu
/r/. 239
Other examples show that longer trills are also possible, especially in slower rates of speech. One example is seen below in Figure 26, produced by speaker LG.
FIGURE 26: Waveform and spectrogram of /paʔatoˈɣorü/
In this example, four closures have been counted during the articulation of the rhotic consonant, and marked with arrows, as in the previous examples. These closures coincide with a decrease in amplitude in the waveform, and lighter shaded regions in the spectrogram, indicating a decrease in acoustic energy as air is briefly obstructed, as the tongue vibrates against the upper surface of the vocal tract. 240
The variability in articulation between a flap (as in Figure 24), and a trill (as in Figures 25 and 26) is likely an example of free variation. I speculate that this variation is possibly tied to rate of speech (although this was not tested), rather than due to external phonological environment. Below, Figures 27 and 28 show two examples of the same word, produced by the same speaker (LG, in each case), where one utterance contains a flap, while the other contains a trill.
FIGURE 27: Waveform and spectrogram of /püraˈvüni/
‘arm’ produced by speaker LG, articulated with a flap. A single closure during the rhotic consonant has been indicated with an arrow. A slight F3 dip is visible.
Here, in Figure 27, only a single flap is visible, which has been marked with an arrow.
The flap itself is very short in duration, typical of flap articulations. The dip in F3 is present, 241
indicating rhoticity, marked with a red line in the figure above. Below, Figure 28 shows the same word produced by the same speaker, with a trilled articulation.
FIGURE 28: Waveform and spectrogram of of /püraˈvüni/
‘arm’ produced by speaker LG, articulated with a trill. Two contacts during the rhotic consonant have been indicated with an arrow.
Both Figure 27 and Figure 28 show the same word, /püraˈvüni/
‘arm’ produced by the same speaker, LG. Only one contact is seen in Figure 27 during the rhotic consonant (a flap), while Figure 28 has two (a trill). It is not the case that flaps or trills are idiosyncratic to each speaker such that one speaker always articulates /r/ as a flap while a different speaker always articulates /r/ as a trill, although some speakers may share a general preference for one or the other. However, the examples in this section have shown a degree of 242
variability: sometimes /r/ is articulated as a flap, and sometimes as a trill. Spectrograms were shown of speakers AP and LG producing both trills and flaps in different words. Additionally, two different utterances of the same word, produced by the same speaker (as in Figures 27 and
28) can have two different rhotic articulations. This illustrates that the flap and the trill exist in free variation in Kawaiisu: sometimes speakers do one, and sometimes the other. Variation in the articulation of rhotic consonants is cross-linguistically common (as reported by Johnson 2008 for
Spanish rhotics, and by Van Bezooijen et al. 2002, Plug and Ogden 2003, etc. , for Dutch). The description of this sound in Zigmond et al. (1990) as a flap, while not incorrect, has been expanded in this section to reflect the variable ways that Kawaiisu speakers actually produce the rhotic consonant /r/.
5.6 The Glottal Stop 5.6.1 Introduction
In Kawaiisu, there are three separate phonological environments where glottal stops may occur. First, glottal stops may occur word-initially before a vowel. Zigmond et al. (1990) analyzes all Kawaiisu words as consonant-initial, such that words which may otherwise would be considered vowel-initial begin with a glottal stop. For example, their dictionary lists the word for
‘raven, crow’ as /ʔatakazi/ rather than /atakazi/. Second, glottal stops may occur intervocalically in word-medial position. In these instances, it is usually the case that the same vowel phoneme occurs on both sides of the glottal stop, as in /moˈʔovi/
There are, however, some instances listed in both Zigmond et al. (1990) and in the Klein 243
recordings (Klein 1958, 1981-1983) where the vowels differ, for example /naraʔüˈsaˈːɣa/
5.6.2 Intervocalic Glottal Stops
There were 210 tokens of words containing glottal stops in word-medial, intervocalic position. Of these, 158 had the same vowel on both sides of the glottal stop, with 52 having different vowels on either side. In these tokens, the glottal stop is often realized as creaky voice on at least one of the adjacent vowels. All four of the Kawaiisu speakers whose speech is examined in this dissertation, with the possible exception of speaker FC, show several instances of words where evidence of creaky voice is clearly visible in the corresponding waveforms and spectrograms. In this section, I show waveforms and spectrograms of Kawaiisu words with intervocalic glottal stops, illustrating a few ways this sound is realized. In more in-depth analyses of phonation, researchers often measure creaky voice by examining spectral tilt (Ladefoged 2003 244
discusses how this is done, see also Gordon and Ladefoged 2001). However, for the purposes of basic illustration, waveforms and spectrograms are sufficient.
Regarding creaky voice, Johnson (2012:170) writes, “in creaky voicing the vocal folds are held together loosely, like two pieces of calf liver, and air bubbles up through them. This results in a longer closed phase of the glottal period and a comparatively shorter open phase.”
This results in glottal pulses being further apart than in modal voicing, and may often be visible as unevenly spaced vertical striations in a spectrogram. Irregularity in spacing between glottal pulses can also be seen in waveforms of creaky voiced sounds, illustrated in some of the following figures. Acoustic and articulatory properties of creaky voice are discussed further in
Gobl 1989, Gordon and Ladefoged 2001, Gerfen and Baker 2005, Avelino 2010, etc. Below,
Figure 29 shows a spectrogram and waveform of /jeˈʔedü/
AP, with visually noticeable creak. 245
FIGURE 29: Waveform and spectrogram of /jeˈʔedü/
Above, in Figure 29, the presence of creaky voice is noticeable. This can be seen in the first portion of the word, up to the second half of the portion marked [eʔe]. Individual glottal pulses can be seen in both the spectrogram and waveform. In the waveform, these are visible as irregularly spaced peaks in amplitude, and in the spectrogram, these can be seen as irregularly spaced vertical striations. By the time the second half of the [eʔe] portion is reached, modal voicing likely continues, as these vertical striations occur closely together, and individual peaks in amplitude are not spaced so far apart in the waveform. Next, Figure 30 shows a waveform and 246
spectrogram of /kotsoɣoˈʔadü/
FIGURE 30: Waveform and spectrogram of /kotsoɣoˈʔadü/
As in Figure 29, the word shown in Figure 30 contains a glottal stop. Here, it is seen between two different vowels in the /oʔa/ sequence. In the portion marked [ʔ], individual pulses can be seen as spaced apart vertical striations in the spectrogram. This can be compared to other vowels in the word, where this spacing is not seen, indicating that the remainder of the vowels in this word likely have modal voicing. 247
In some instances, the waveforms or spectrograms do not show clear creaky voice. In such cases, where a /VʔV/ sequence is present, there are other ways the glottal stop may be realized. One such example was shown in the previous section on rhotic trills, in Figure 26. In that example, /paʔatoˈɣorü/
FIGURE 31: Waveform and spectrogram of /viˈnoʔo/
248
In this example, creaky voice is not apparent based on the information seen in the spectrogram or the waveform. However, immediately following the /n/, formants for /o/ become visible. For a brief period of time (about 40ms) it can be seen in the waveform that amplitude during this /o/ is similar to the preceding sounds. Then, there is a sudden decrease in amplitude, which is visible in the waveform, and the dark formant bands become lighter. This decrease in amplitude corresponds to the glottal stop, which in this example lasts for a duration of approximately 45ms, before amplitude increases again, indicating the final portion of the /oʔo/ sequence. Here, it appears to be the case that /ʔ/ is articulated as an approximant with quality interpolated between the two surrounding vowels.
5.6.3 Pre-glottalized Sonorants
In the Zigmond et al. (1990) dictionary, there are several words which appear to have internal sequences of the glottal stop followed by a sonorant. In total, there were 59 instances of phonological glottal stop plus sonorant sequences noted in the Klein data. One such word which frequently occurs in the Klein recordings (Klein 1958, 1981-1983) is /taʔniˈpüzi/
‘man’. Glottal stops in this position tend to surface in a similar way to glottal stops in intervocalic position, discussed in the previous subsection. In many instances, pre-glottalized sonorants are associated with the presence of creaky voice on the immediately preceding vowel.
One such example is shown below in Figure 32. 249
FIGURE 32: Waveform and spectrogram of /taʔniˈpüzi/
Above, it can be seen that creaky voice occurs in the vicinity of the glottal stop in this utterance of /taʔniˈpüzi/
(AP) used creaky voice almost exclusively in this environment, but another speaker (FC) used modal voice almost exclusively in the same context. Speaker LG varied between creaky and 250
modal voicing, while speaker RB only produced one token of this sound, in which creaky voice was not apparent. One example of this sound without evidence of creaky voice is shown below, in Figure 33.
FIGURE 33: Waveform and spectrogram of /taʔniˈpüzi/
The example shown in Figure 33 is in contrast with the example shown in Figure 32. In
Figure 32, evidence of creaky voice was visible, but in Figure 33, no such evidence is present, suggesting modal voicing throughout. In fact, there appears to be no evidence for a glottal stop at all in Figure 33. In words containing the sequence of glottal stop plus sonorant, such as these, both articulations are common (although some speakers may have a general preference for one 251
versus the other), suggesting some degree of variability: sometimes the glottal stop surfaces as creaky voice in these instances, and sometimes it appears to be absent.
5.7 Pre-aspirated Sonorants In addition to word-medial glottal stop plus sonorant sequences, Zigmond et al. (1990) also lists several instances where sonorant consonants occur word-medially immediately following /h/. One such example is /pohˈnija/
FIGURE 34: Waveform and spectrogram of /pohˈnija/
Above, Figure 34 shows a spectrogram and waveform of /pohˈnija/
Another example is shown below. 253
FIGURE 35: Waveform and spectrogram of /tahˈmana/
As in Figure 34, glottal pulses have been made visible in the waveform for Figure 35.
This is the form /tahˈmana/
5.8 Conclusion and Discussion In this chapter, I have presented acoustic analyses and examples of several facets of the consonant system of Kawaiisu. The consonants that were referred to as velar obstruents /g/ and
/gw/ in past literature were shown to frequently possess acoustic characteristics more typical of fricatives or approximants than true stop consonants. Occlusion was only apparent in approximately 34% of tokens, and only 12% had visible bursts. Thus, a large majority of these sounds were more commonly articulated as fricatives or approximants. This leads to an asymmetry in the stop series of Kawaiisu. The voiceless stops /p, t, k, kw/ are present in the phoneme inventory, but there are only two stops in the voiced series: /b, d/. Phonological asymmetries such as this are well-attested in the languages of the world, and there are aerodynamic explanations for why a language may develop in this way, discussed, for example, by Sherman (1975), Ohala & Riordan (1979), and Ohala (1983, 1997).
I have also examined voice onset time (VOT) of voiceless Kawaiisu stops. The findings on VOT are consistent in many ways with established literature. For instance, Kawaiisu stops articulated toward the front of the vocal tract tend to be associated with a shorter VOT duration than those at more posterior places of articulation. This is in line with the findings of Fischer-
Jørgensen 1954, Peterson and Lehiste 1960, and Cho and Ladefoged 1999 who note similar findings cross-linguistically. Some other factors, such as level of stress and word position, did not seem to have an effect on VOT in Kawaiisu. Length of the following vowel, however, does have an effect on VOT duration in Kawaiisu such that when a voiceless stop is followed by a phonemically long vowel, VOT duration is greater than when the stop is followed by a short vowel. A similar pattern can be found in Finnish, which has a contrast between long and short vowels. In Finnish, VOT tends to be greater when a surrounding vowel is phonemically long 255
(Doty et al. 2007). The relationship between VOT and phonemic vowel length in additional languages with such a contrast is a worthy topic of research.
Additionally, the findings presented in this chapter provide support for the description made by Zigmond et al. (1990) of the process of palatalization. Through their prose description and glossed examples, they suggest that coronal fricative and affricate phonemes in Kawaiisu are palatalized when immediately following a front vowel /i/ or /e/. The examples shown in this chapter provide acoustic evidence for these claims. The examples shown in this chapter indicate that acoustic energy for the Kawaiisu phonemes /s/and /ts/ is concentrated in a lower frequency range when these phonemes are immediately preceded by front vowels. In other environments the concentration of acoustic energy was shown to lie in a somewhat higher frequency range.
This suggests that Zigmond et al’s (1990) description of palatalization in Kawaiisu is correct.
This chapter also examines the rhotic /r/, finding some degree of variability in the ways this consonant is pronounced. In some cases, it occurs as a trill, with multiple strikes of the oral articulators in rapid succession, and in other cases, it occurs as a flap, with only one quick contact of the articulators. This does not appear to be phonologically conditioned in Kawaiisu.
Finally, aspects of the glottal phonemes /ʔ/ and /h/ were examined. The glottal stop is often realized as creaky voice in intervocalic position, as well as when preceding a sonorant consonant. When /h/ precedes a sonorant consonant, the result is often a period of voicelessness before the sonorant.
Phonologically, the existence of these sequences (glottal plus sonorant, and /h/ plus sonorant) raises some concerns. Regarding word-internal consonant clusters, Zigmond et al.
(1990:6) write, “consonant clusters never occur initially in native words, but they are fairly 256
common internally. Some possible surface clusters include -mb-, -nb-, -sk-, -skw-, -hw-, -hn-, -
ʔn-, -nt-, and -nd-.” Elsewhere in the text, they note that prenasalized forms no longer exist, and are a feature of an earlier stage of the language. In their dictionary, almost all words consist of
CV syllables only, casting doubt on their claim that word-internal consonant clusters are “fairly common.” When words are uttered with two word-internal adjacent consonants, which occasionally happens, this is usually the result of word-internal vowel deletion, examples of which are shown in Chapter 3 of this dissertation. In the Klein recordings (Klein 1958, 1981-
1983), the majority of words which seem to have internal consonant clusters not resulting from word-internal vowel deletion are those containing the suffix /-vistü/ ‘a lot’, as in /ʔabiɣiˈvistü/
If it is the case that Kawaiisu has CV as its default syllable shape, then the existence of forms like /taʔniˈpüzi/
Clements and Keyser (1983) and Ito (1987) on syllabification of onsets and codas). Assuming
CVC syllables are disfavored in Kawaiisu, there are at least two possible explanations for this discrepancy. First, if these phonemes are analyzed as complex segments rather than sequences of two discrete segments (for example, as a single / hn/ rather than a sequence of /h/ followed by
/n/), then CVC can be removed from the list of legal syllable types in Kawaiisu (with the rare exception for morphemes like /-vistü/, discussed above). In this case, a word like /pohˈnija/ 257
On the other hand, it is well-attested that languages which otherwise lack codas may sometimes allow them in certain conditions. One fairly common cross-linguistic restriction on codas is that they may not have consonantal place features (see Ito 1989, Beckman 2004, etc.).
Phonemes like /h/ and /ʔ/ are unique in that they are often argued to lack consonantal place of articulation features, evidenced in part by the fact that the articulation of such sounds does not require any gesture from an oral articulator (Sagey 1986, Bessell 1992, Boroff 2007). Thus, these phonemes may often serve as codas in languages which otherwise disfavor codas, meaning that an analysis of a Kawaiisu word like /pohˈnija/
A similar situation can be found in Yurok. Yurok has a phonemic distinction between plain and glottalized sonorants (Blevins 2003). Like Kawaiisu, glottalization of sonorants is often realized as creaky voice on a preceding vowel. Similarly, the distinction between plain and glottalized sonorants is only maintained in certain positions, namely in word-medial and word- final position. Word-initially in Yurok, only plain sonorants are found, as in Kawaiisu. Blevins
(2003) writes that, in Yurok, certain phonological processes require glottalized sonorants to be analyzed as single segments, such as /ʔm/, /ʔw/, etc. However, in intervocalic position they are often analyzed as clusters, with the glottal stop serving as a coda for one syllable, and a sonorant as the onset for the next. Evidence for this can be seen in the stress system of Yurok, which depends on syllable weight. Closing a syllable with a glottal stop coda has the effect of making a syllable heavy, which affects the placement of stress. Further, Blevins (2003) notes that Yurok 258
speakers, when asked for judgments regarding syllable boundaries, consistently break up glottalized sonorants such that there is a syllable break between a glottal stop and a following consonant, analyzing these as clusters, rather than complex segments. In other cases, these judgments are often inconsistent between different speakers.
Determining whether these should be analyzed as complex segments or as discrete sequences in Kawaiisu is a question of phonology, rather than phonetics, and may lie somewhat outside the scope of this dissertation. The instrumental techniques discussed here cannot easily be used to precisely answer this phonological question. Like the Yurok speakers discussed in
Blevins (2003), Kawaiisu speakers may have intuitions regarding whether these represent single sounds or a sequence of sounds, but in a language with few remaining native L1 speakers, such information may be difficult to collect. Additionally, phonological evidence for analyzing glottalized or aspirated sonorants as either bi- or mono-segmental, as used by Blevins (2003) for
Yurok, is not present in Kawaiisu.
Finally, there is a question of whether the glottal stop should be analyzed independently as a consonant phoneme, or simply an additional laryngeal feature belonging to the vowel. In
Coatzsopan Mixtec, for example, it has been argued that the phoneme inventory does not contain a glottal stop, but rather that certain vowels are simply laryngealized (Macaulay & Salmons
1995, Gerfen 1999, Gerfen & Baker 2005). In Coatzsopan Mixtec, coda consonants generally do not occur. However, in some cases it appears as if the glottal stop can appear in coda position in this language. As in Kawaiisu, this may sometimes surface as laryngealization or creaky voice on an adjacent vowel. Gerfen (1999) and Gerfen & Baker (2005) argue that, in Coatzsopan Mixtec, this laryngealization should be analyzed as a feature belonging to the vowel itself, rather than 259
positing a separate phoneme which only apparently occurs in coda position in a language where coda consonants are generally disallowed.
Another case where laryngeals display similar behavior can be found in the Desano language (Silva 2016). In this language, the only codas that are permissible are the glottal segments [h] and [ʔ]. While these sounds are associated with a set of restrictions in both Desano and Kawaiisu, the restrictions that apply to these segments are quite different. In Desano, Silva
(2016) points out that these segments only occur in roots, post-vocalically (intervocalically and preconsonantally), and that they never occur in affixes. Thus, Silva (2016) analyzes these glottal segments not as full consonants, but as the realization of a laryngeal feature occurring after the first vowel in a root. In Kawaiisu, the glottal segments can either occur as onsets, or as apparent codas, but are not restricted to roots. These segments often occur in Kawaiisu affixes, and as such an analysis which posits the laryngeal consonants as features tied to roots would not be favored in the case of Kawaiisu.
Like Coatzsopan Mixtec and Desano, coda consonants are generally disfavored in
Kawaiisu, as mentioned above in the discussion of glottal plus sonorant sequences. Generally, it is unclear whether these should be analyzed as sequences or as complex segments. In intervocalic position however, there is evidence for analyzing the glottal stop as an independent phoneme, rather than assuming laryngealization to be a feature of adjacent vowels. In Kawaiisu, stress almost always falls on the penultimate syllable, except in cases of word-final vowel deletion, or if the final vowel in the word is long, as discussed in Chapters 1 and 4 of this dissertation. Chapter 4 revealed that stress in Kawaiisu is associated with an increase in pitch: vowels which are stressed have higher pitch than unstressed vowels. We also find that several 260
Kawaiisu words exist where the glottal stop apparently exists as an onset, creating a new syllable for purposes of stress assignment. One such example is /kaˈʔadü/
261
CHAPTER 6: CONCLUSION
6.1 Introduction
Within the realm of work dealing with indigenous American languages, Kawaiisu is a comparatively understudied language. A lexicon and general guide to the grammar exists
(Zigmond et al. 1990), and some articles have been published dealing with specific aspects of morphology (Booth 1979) and phonology (Klein 2002). The present dissertation represents the first in-depth analysis and description of the phonetic structures of the language, using instrumental methods to achieve this goal. This work grew from the analysis of Thomas (2017), in which the Praat software (Boersma and Weenink 2017) was used to analyze the stressed vowels of Kawaiisu, with a particular eye on the vowel labeled in this dissertation as /ü/. The present dissertation is an expansion of this earlier work on Kawaiisu vowel phonetics, describing the entire vowel space, rather than just focusing on a single vowel phoneme. The analyses presented here discuss long and short vowels, in both stressed and unstressed positions. Phonetic aspects of stressed and unstressed vowels were compared to each other, in an attempt to determine and describe the acoustic correlates of Kawaiisu stress. Additionally, I examine various aspects of the consonantal system of Kawaiisu, determining which factors influence the duration of voice onset time in Kawaiisu voiceless stops, and illustrating articulatory and acoustic variability of several other Kawaiisu consonants. In the concluding chapter of this dissertation, I briefly summarize the key findings of previous chapters, discuss the implications of this work for Kawaiisu community members and academics working on endangered languages, and identify some possible questions for future research. 262
6.2 Summary of Previous Chapters
In the introductory chapter, I discuss the state of Kawaiisu language endangerment and steps the community has taken towards language revitalization. Information regarding their language revitalization activities can be found at the Kawaiisu Language & Culture website, accessible at http://www.kawaiisu.org (Kawaiisu Language & Cultural Center 2018). I note that
Kawaiisu is classified as a Stage 8b language in Lewis & Simons’s (2010) Extended Graded
Intergenerational Disruption Scale, due to the small number of fluent L1 speakers who learned the language as children (Green 2013). At this level of language endangerment, Lewis & Simons
(2010:112) write, “the only remaining speakers are among the grandparent or great grandparent generation, and are so few or so scattered that they have little opportunity to use the language with each other.” Steps have been taken towards revitalization however, which are outlined in
Chapter 1 in this dissertation. At the individual and family level, community members are encouraged to speak the language at home when possible, stressing the ways in which modern digital technology such as smartphones, tablets, and computers can be used in language learning
(and making recordings of the language which can be played back and studied at any time).
Community members and linguists have come together to create a practical grammar, accessible to the public and written with minimal technical linguistics jargon, for the use of community members who may not have a background in linguistics.
At a broader, community-based level, projects have been designed to aid community members in developing teaching material, and recording conversations among speakers. The
Kawaiisu Language & Cultural Center also worked on projects with the goal of developing
Master-Apprentice programs designed to immerse families and individual language learners in 263
the Kawaiisu language. DVDs and CDs have also been produced which educate community members on language and traditional practices.
In this chapter, I also discuss the genetic affiliation of Kawaiisu, noting that it is a Uto-
Aztecan language, belonging to the Central Numic branch. Its closest linguistic relatives include
Comanche, Shoshoni, and Timbisha. A large portion of the remainder of the chapter provides a brief overview of some grammatical structures of Kawaiisu, summarized from Zigmond et al.
(1990).
Chapters 2 and 3 present discussions of the stressed and unstressed vowel spaces of
Kawaiisu, respectively. In these chapters, I find that differences in duration exist between phonemically long and phonemically short vowels, both stressed and unstressed, with the phonemically long vowels having greater duration. Further, long and short vowels are compared to each other, testing for differences in quality. These tests reveal that phonemically long vowels tend to occupy a more peripheral area in the vowel space than phonemically short vowels.
In addition to tests on duration and peripherality of vowel quality between long and short vowel pairs, it is shown in Chapter 3 that vowels in Kawaiisu are often deleted. This process is shown to be optional in Kawaiisu, and may be tied to rate of speech. Though I do illustrate examples of vowels being deleted in word-internal position, I pay specific attention to vowels in word-final position, which Zigmond et al. (1990:146) discuss as being particularly susceptible to deletion, noting further that vowel deletion in Kawaiisu “seems to be a correlate of the consistent devoicing or deletion of final vowels in other Southern Numic languages.” While they are correct that vowels in Kawaiisu are often dropped word-finally, they argue that this happens 264
most often in verbs. In the Klein data which form the basis of the acoustic analyses presented here, I note that word-final vowel deletion does not necessarily favor one lexical category over any other, and suggest that word-final vowel deletion happens more often in utterance-final position. As Zigmond et al. (1990:15) note, Kawaiisu is “dominantly SOV,” which leads to a situation where verbs are likely to end sentences (although many other word orders are possible in Kawaiisu). If Kawaiisu is dominantly SOV, and the end of an utterance aligns with the end of a sentence, and sentences tend to be verb-final, then this may explain why Zigmond et al. (1990) coincidentally find that word-final vowel deletion often affects verbs. However, evidence for word-final vowel deletion affecting verbs specifically, outside of utterance-final contexts, was not discovered to be present in the Klein wordlist elicitation recordings (Klein 1958, 1981-1983).
In Chapter 4, I compare the analyses of stressed and unstressed Kawaiisu vowel spaces, seeing how they differ, in an attempt to describe the acoustic correlates of Kawaiisu lexical stress. Aside from vowel quality, I focus on duration and pitch, finding that pitch is a reliable indicator of stress. Stressed vowels in Kawaiisu have higher pitch than unstressed ones. The relationship between stress and vowel duration was not entirely consistent, with results indicating stress plays a strong role in the duration of the /u/ and /o/ vowel qualities (with greater duration in stressed syllables), but not necessarily in other vowels. The lack of a solid pattern regarding vowel duration and stress in Kawaiisu is consistent with Berinstein (1979), who argues that if a language uses a suprasegmental feature (like duration) in a phonemic contrast, then that feature is less likely to be used as a strong indicator of stress. Kawaiisu has both long and short vowels, so these findings were expected. Regarding differences in vowel quality, I find that unstressed Kawaiisu vowels sometimes occupy a more peripheral position in the vowel space 265
than stressed vowels. This is somewhat unexpected, given the usual cross-linguistic tendency for unstressed vowels to be more centralized and stressed vowels to be more peripheral (Tiffany
1959). I argue that this unexpected pattern may arise due to the relationship (or lack thereof) between stress and vowel duration in Kawaiisu. Generally, however, patterns of greater peripherality for unstressed vowels are inconsistent, but it is clear at least that unstressed vowels are not reduced in the same way that vowels in other languages may often be reduced. Both stressed and unstressed vowels occur in relatively peripheral positions in the vowel space.
Finally, Chapter 5 provides an analysis and description of Kawaiisu consonants. In this section, I use acoustic evidence to show that much variation exists in the articulation of Kawaiisu consonants. I find that the phoneme labeled as /g/ by Zigmond et al. (1990) is articulated as a fricative or approximant in the majority of utterances, but can also be pronounced as a plosive, with a visible release burst in a spectrogram. I also provide support for the claim made by
Zigmond et al. (1990) that the coronals /s/ and /ts/ are palatalized when they are immediately preceded by a front vowel. This is apparent due to the greater concentration of acoustic energy at lower frequencies (a likely indicator of a more posterior articulation, as discussed by Jongman et al. (2000) and Ladefoged (2003)) during the articulation of /s/ and /ts/ in this phonological environment. I also show that the rhotic /r/ can be pronounced with either a flap articulation or as a trill, and suggest that this variation may be tied to rate of speech. Additionally, examples of the glottal stop are shown in different environments, illustrating that the glottal stop is often associated with the presence of creaky voice. Examples of /h/ preceding a nasal are also shown, illustrating a duration of voicelesness preceding the nasal, and phonological questions are raised concerning the phonemic status of sequences like /hn/ and /ʔn/. Lastly, this chapter investigates 266
the duration of voice onset time for voiceless stops in Kawaiisu. I find that consonants with a more posterior articulation are associated with greater VOT duration, which is expected, given the findings of others (for example, Cho and Ladefoged 1999). I also illustrate that voiceless stops in Kawaiisu have longer VOT when followed by a phonemically long vowel.
6.3 Implications for the Community
From a practical standpoint, this dissertation may be useful to both academics working with severely endangered languages and to members of the Kawaiisu community engaged in the revitalization effort. On the academic side, this dissertation highlights the usefulness of archival audio data in linguistic description and documentation. Kawaiisu is but one of many languages with such audio documentation available online in the California Language Archives. Like
Kawaiisu, many of these languages are also underdocumented, understudied, and endangered.
Some may be completely dormant, with no speakers remaining. In these cases, archival data may be all that is left of a once widely-spoken language. This dissertation shows that important descriptive and documentary work can still be done. This dissertation also highlights some of the difficulties of working with such archival data. Other studies involving phonetic description of indigenous languages based on archival recordings include Tucker (2007) on Chemehuevi, a
Southern Numic language closely related to Kawaiisu, and Lawyer (2015) on Patwin. In working with such data, audio quality may be poor or inconsistent due to very old recording equipment.
The researcher undertaking the recordings may not have phonetic description as a primary goal, resulting in wordlist elicitations that are inconsistent across speakers or not ideal for this type of work. The linguistic data may sometimes be obscured by children playing, people coming and going, household appliances buzzing, tapes hissing, people talking over each other, birds 267
chirping, poor microphone placement, and television or radio playing in the background.
Examples of all these things can be found in the Klein recordings (Klein 1958, 1981-1983).
Nevertheless, in the midst of such extraneous factors, clear, useful data can be found in archival sources, which would otherwise be lost. In the case of severe language endangerment, the best data are the available data.
From a revitalization perspective, this dissertation provides a thorough description of the sounds of the Kawaiisu language. It does not necessarily show a correct or proper way to pronounce the sounds of Kawaiisu like a native speaker, but shows that the language is rich with variation: vowels can often be deleted in some contexts, some sounds are palatalized in other contexts, sounds like /ɣ/ and /r/ can vary freely between different manners of articulation, etc . In any language, there is often no single, correct way. Kawaiisu is no exception.
This dissertation itself is not written to serve directly as revitalization material for use in the community. However, a community linguist can use this information to help in the creation of pedagogical materials, discussing the different ways Kawaiisu sounds may be pronounced, supplementing existing materials. One must also keep in mind that the descriptions in this dissertation come from recordings made (at the time of this writing) roughly 35 – 60 years ago
(Klein 1958, 1981-1983). All languages are affected by change over time, including changes to sound systems and pronunciations. This is especially true of severely endangered languages, where change may occur more rapidly, and in such cases it may be very difficult to predict the ways in which changes may occur (Campbell and Muntzel 1989). Thus, it is possible that a community member or linguist working with the community may notice some differences 268
between the descriptions presented in this document, and the ways in which current L1 speakers continue to use the language. This does not mean current speakers of the language are incorrect, nor does it necessarily diminish the veracity of the descriptions and analyses presented here.
Such differences, if they are found, may be a result of different dialects, different idiolects, or due to the inevitability of change which all languages undergo.
6.4 Questions for Future Research
Throughout this dissertation I have attempted to point out possible questions for future research. This includes the study of VOT in additional languages which maintain a phonemic contrast in vowel length, and an investigation of the status of Kawaiisu sequences where a glottal stop or /h/ immediately precede word-internal sonorant consonants, both from the perspective of synchronic phonology, and of their diachronic development. One possible area of study that relates to Kawaiisu which has not been mentioned involves the narratives presented in Zigmond et al. (1990). Two narrative stories are presented in the Zigmond et al. (1990) text, with complete glosses. These texts may be examined for the possible discovery of grammatical or semantic patterns which have not been described elsewhere. Further, some tapes in the Klein recordings
(Klein 1958, 1981-1983) contain conversations and stories in Kawaiisu which can still be examined in more detail with the aid of native speakers. When dealing with such an underdocumented language, it may be impossible to examine faster rates of speech without assistance from a speaker who is familiar with the language. Examining natural speech is important, as it may contain some features which are not present in more careful rates of speech, as in the wordlist elicitation sessions examined in this dissertation. 269
6.5 Conclusion In this final section, several of the findings of this dissertation are briefly summarized.
First, Kawaiisu has been shown to have six distinct phonemic vowel qualities, with two phonemic lengths. For the most part, the primary difference between phonemic vowel length is duration: phonemically long vowels tend to be held for a greater period of time than phonemically short vowels. Long and short vowels may differ in slightly quality for some speakers, for some vowels, but no consistent patterns were found. If one speaker has a certain vowel which differs in quality depending on phonemic length, other speakers may have a different vowel which shows differences in quality. When vowels differed, the sets of vowels exhibiting changes in vowel quality were often different depending on stress. For instance, if one speaker’s /a/ vowel shows differences in vowel quality based on phonemic length in stressed syllables, that is not a guarantee that this speaker’s /a/ vowel will show the same difference in unstressed syllables. Thus, while some speakers may have some vowels that differ in quality depending on whether or not they are phonemically long or phonemically short, those differences may be realized differently for different speakers, or even for the same speaker, when considering opposite levels of stress.
On Kawaiisu stress, it was found that pitch is a much more important acoustic cue than duration. Kawaiisu syllables have higher pitch when they carry stress. They may or may not be longer. Further, Kawaiisu unstressed vowels do not seem to undergo reduction or centralization in the same way that English vowels do. Rather, unstressed Kawaiisu vowels are characterized by a strong lack of reduction, sometimes even occurring closer to the vowel space periphery than stressed vowels. 270
Consonants were shown to be highly variable in articulation, with multiple articulations commonly occurring for several phonemes, such as /ɣ/, /r/, /s/, and /ts/. While the different articulations for /s/ and /ts/ seem to be part of an assimilatory process in which these phonemes become palatalized when adjacent to front vowels, the different articulations for /ɣ/ and /r/ are likely examples of free variation. The /ɣ/ phoneme is commonly produced with fricative or approximant-like articulation, and rarely as a fully occluded stop. The /r/ phoneme is pronounced variably as a flap or trill. These variable pronunciations do not appear to be phonologically conditioned. Finally, VOT is somewhat variable for voiceless Kawaiisu stops: VOT is greater for consonants with posterior articulations, and when the stop occurs in a syllable which also contains a phonemically long vowel.
This work represents the first large scale instrumental study of the phonetics of Kawaiisu, an endangered and underdocumented language of the Americas. Although this work uses pre- recorded archival audio data, and it is preferable to work in person, with living speakers, any data is good data when dealing with a severely endangered language. Numerous languages exist where archival data, like the kind examined here, is available. In some cases, archival data may be all that is available. By continuing to examine archival data and describe underdocumented indigenous languages, linguists can help provide opportunities to enrich indigenous language revitalization programs, and can advance the field of linguistics, making new discoveries. 271
APPENDIX: LIST OF TOKENS This appendix contains a list of the words taken from the Sheldon Klein Collection of
Kawaiisu Sound recordings which were selected for analysis in this dissertation (Klein 1958,
1981-1983). Only those words which were selected for analysis are contained in this list.
Speakers are referenced by their initials, as in the remainder of the dissertation. Under the
“archival item number” heading, I list the name of the file where each specific token can be found in the California Language Archive. The third column is a phonemic transcription of the
Kawaiisu word, given in the International Phonetic Alphabet. In the fourth column, I attempt to transcribe each word in the orthography adopted by the Kawaiisu community, an overview of which can be found in Chapter 1 of this dissertation, and on the Kawaiisu Language and Cultural
Center website, at http://www.kawaiisu.org (Kawaiisu Language & Cultural Center 2018).
Glosses for each individual word can be found in the Zigmond et al. (1990) lexicon and in the
Klein recordings (Klein 1958, 1981-1983). Because the Kawaiisu orthography is not standardized, some speakers may spell some words differently. Any errors in orthographic or phonemic transcription are my own. The list itself is organized by speaker, in alphabetical order by first name (AP, FC, LG, RB), then in numerical order by archival item number.
Archival Item Speaker Phonemic Transcription Number Orthographic Transcription
LA65.067 AP /taːsuˈʔuvi/ taasu’uvi
LA65.067 AP /ʔaɣaˈkidü/ aaxkid
LA65.067 AP /wiziˈkizi/ wizikizi 272
LA65.067 AP /ʔeˈwutsi/ ewutsi
LA65.067 AP /ˈnüʔü/ nü’ü
LA65.067 AP /kaʔaˈvaːdü/ ka’avaadü
LA65.067 AP /ʔaˈbiɣi/ abigi
LA65.067 AP /ʔaˈbiɣi/ abigi
LA65.067 AP /jühuˈɣadü/ yühugadü
LA65.067 AP /jühuˈvüdü/ yühugadü
LA65.067 AP /nüwüˈʔami/ nüwü’ami
LA65.067 AP /nüwüˈʔami/ nüwü’ami
LA65.067 AP /taʔnipüˈzija/ tanipüziya
LA65.067 AP /hüʔükweːˈvaːdü/ hü’ükweevaadü
LA65.067 AP /haˈvamü/ havamü
LA65.067 AP /kaˈʔapü/ ka’apü
LA65.067 AP /aˈwatü/ awat
LA65.067 AP /haˈvidü/ havidü
LA65.067 AP /haviˈnübü/ havinübü
LA65.067 AP /saˈʔampü/ sa’ampü
LA65.067 AP /ˈʔaja/ aya
LA65.067 AP /toveʔeˈpiːtsi/ tove’epich
LA65.067 AP /ˈʔaja/ aya
LA65.067 AP /hoˈjovi/ hoyovi 273
LA65.067 AP /ʔohˈnidü/ ohnidü
LA65.067 AP /ˈʔohni/ ohni
LA65.067 AP /taʔniˈpüzi/ tanipüz
LA65.067 AP /piːvaniˈjadü/ piivaniyad
LA65.067 AP /puˈɣuzi/ puguzi
LA65.067 AP /tsüɣüˈpizi/ tsügüpiz
LA65.067 AP /puˈɣuzi/ puguzi
LA65.067 AP /tsüɣüˈpizi/ tsügüpiz
LA65.067 AP /süniˈʔarü/ süni’arü
LA65.067 AP /kukopi/ kukop
LA65.067 AP /sünaˈʔavi/ süna’av
LA65.067 AP /ˈtiːpü/ tiipü
LA65.067 AP /maˈhavü/ mahavü
LA65.067 AP /ʔoˈwavi/ owavi
LA65.067 AP /sünaˈʔavi/ süna’av
LA65.067 AP /ʔaˈwatü/ awat
LA65.067 AP /sünaˈʔavi/ süna’av
LA65.067 AP /püˈkeːdü/ pükeed
LA65.067 AP /ˈnüʔü/ nü’ü
LA65.067 AP /tsüɣüˈpizi/ tsügüpiz
LA65.067 AP /tsüɣüˈpizi/ tsügüpiz 274
LA65.067 AP /ʔoˈhobü/ ohobü
LA65.067 AP /ʔohoˈvümü/ ohovüm
LA65.067 AP /puˈɣuzi/ puguzi
LA65.067 AP /hüˈvivi/ hüvivi
LA65.067 AP /ˈtiːpü/ tiipü
LA65.067 AP /totsivaˈʔaːmi/ totsiva’aami
LA65.067 AP /muˈhutsi/ muhutsi
LA65.067 AP /ˈʔuvas/ uvas
LA65.067 AP /ˈpüːpü/ püüpü
LA65.067 AP /kuˈtsapü/ kutsap
LA65.067 AP /kutsaˈwaːka/ kutsawaaka
LA65.067 AP /ˈsiːni/ siini
LA65.067 AP /naˈɣeːka/ nageeka
LA65.067 AP /juʔuˈvinü/ yu’uvinü
LA65.067 AP /soːˈkidü/ sookid
LA65.067 AP /ˈsoːpü/ soopü
LA65.067 AP /poˈʔodü/ po’odü
LA65.067 AP /poˈʔodü/ po’odü
LA65.067 AP /kaˈʔadü/ ka’adü
LA65.067 AP /kaˈʔadü/ ka’adü
LA65.067 AP /taviˈnübü/ tavinübü 275
LA65.067 AP /karüˈnübü/ karünübü
LA65.067 AP /laˈmeːsa/ lameesa
LA65.067 AP /kaˈnaːka/ kanaaka
LA65.067 AP /kaˈʔapü/ ka’apü
LA65.067 AP /neheˈkapü/ nehekap
LA65.067 AP /würeːˈwübü/ würeewübü
LA65.067 AP /poˈɣwitü/ pogwit
LA65.067 AP /poˈɣwitü/ pogwit
LA65.067 AP /pijaˈviːna/ piyaviina
LA65.067 AP /poˈɣwitü/ pogwit
LA65.067 AP /poˈɣwitü/ pogwit
LA65.067 AP /pijaˈviːna/ piyaviina
LA65.067 AP /woʔoˈravi/ wo’orav
LA65.067 AP /paˈtseːna/ patseena
LA65.067 AP /laˈmeːsa/ lameesa
LA65.067 AP /ˈkeːvi/ keevi
LA65.067 AP /ˈkahni/ kahni
LA65.067 AP /nohokweːˈneːna/ nohokweeneena
LA65.067 AP /pükeːˈdiːka/ pükeediika
LA65.067 AP /saˈpiːna/ sapiina
LA65.067 AP /taˈwami/ tawami 276
LA65.067 AP /püˈkeːdü/ pükeed
LA65.067 AP /ʔeːˈpizi/ eepizh
LA65.067 AP /ʔeːˈpizi/ eepizh
LA65.067 AP /ˈkeːvi/ keevi
LA65.067 AP /ˈkeːvi/ keevi
LA65.067 AP /seːˈɣidü/ seegid
LA65.067 AP /pükeːküˈdüni/ pükeeküdüni
LA65.067 AP /wahaju/ wahayu
LA65.067 AP /seːˈɣidü/ seegid
LA65.067 AP /pükeːküˈdüni/ pükeeküdüni
LA65.067 AP /seːˈɣidü/ seegid
LA65.067 AP /ʔeːˈpizi/ eepizh
LA65.067 AP /nukwiˈdümü/ Nukwidümü
LA65.067 AP /ʔeːˈpizi/ eepizh
LA65.067 AP /pükeːˈküdü/ pükeeküdü
LA65.067 AP /puˈɣuzi/ puguzi
LA65.067 AP /püˈkeːdü/ pükeed
LA65.067 AP /taˈwiʔi/ tawi’i
LA65.067 AP /kahniˈɣadü/ kahnigad
LA65.067 AP /kahniˈɣadü/ kahnigad
LA65.067 AP /nuˈkwidü/ nukwidü 277
LA65.067 AP /nuˈkwidü/ nukwidü
LA65.067 AP /kaˈkarü/ kakarü
LA65.067 AP /kaˈkarü/ kakarü
LA65.067 AP /kaˈkarü/ kakarü
muhni LA65.069 AP /ˈmuhni/
LA65.069 AP /tukuːˈmüːtsi/ tukuumüüts
LA65.069 AP /sünaˈʔavi/ süna’av
LA65.069 AP /ˈpuʔi/ pu’i
LA65.069 AP /puˈʔivi/ pu’ivi
LA65.069 AP /puˈʔivi/ pu’ivi
LA65.069 AP /pijaˈviːna/ piyaviina
LA65.069 AP /pijaˈviːna/ piyaviina
LA65.069 AP /ˈtiːpü/ tiipü
LA65.069 AP /ˈpüːpü/ püüpü
LA65.069 AP /ˈneːdü/ needü
LA65.069 AP /ˈnüwü/ nüwü
LA65.069 AP /nüwüˈkahni/ nüwükahni
LA65.069 AP /ˈtiːpü/ tiipü
LA65.069 AP /naˈroʔo/ naro’o
LA65.069 AP /naˈroʔo/ naro’o 278
LA65.069 AP /ʔeːˈpizi/ eepizh
LA65.069 AP /ʔeːˈpitsi/ eepich
LA65.069 AP /ʔeːˈpitsi/ eepich
LA65.069 AP /taˈnimü/ tanimü
LA65.069 AP /taˈnimü/ tanimü
LA65.069 AP /ˈneːzi/ neezh
LA65.069 AP /toˈɣowa/ togowa
LA65.069 AP /sünaˈʔavi/ süna’av
LA65.069 AP /sünaˈʔavi/ süna’av
LA65.069 AP /kaʔaˈdünü/ ka’adün
LA65.069 AP /pijaˈviːna/ piyaviina
LA65.069 AP /ˈsuːju/ suuyu
LA65.069 AP /pijaˈviːna/ piyaviina
LA65.069 AP /ʔabiɣiˈvistü/ abigivishtü
LA65.069 AP /ˈkahni/ kahni
LA65.069 AP /paˈjeːka/ payeeka
LA65.069 AP /paˈjeːka/ payeeka
LA65.069 AP /kahˈnija/ kahniya
LA65.069 AP /taʔnipüˈzija/ tanipüziya
LA65.069 AP /moˈmoʔo/ momo’o
LA65.069 AP /taʔniˈpüzi/ tanipüz 279
LA65.069 AP /nüwükahˈniːna/ nüwükahniina
LA65.069 AP /ˈkahni/ kahni
LA65.069 AP /ˈnüːwü/ nüwü
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /tüˈhüja/ tühüya
LA65.069 AP /tüˈhüja/ tühüya
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /ˈtiːpü/ tiipü
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /ˈtiːpü/ tiipü
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /ˈtübi/ tübi
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /laˈmeːsa/ lameesa
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /woʔoˈravi/ wo’orav
LA65.069 AP /woʔoˈravi/ wo’orav 280
LA65.069 AP /ˈpoʔo/ po’o
LA65.069 AP /woʔoˈravi/ wo’orav
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /woʔoˈravi/ wo’orav
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /üveˈpitsi/ üvepich
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /karüˈnübü/ karünübü
LA65.069 AP /ˈkahni/ kahni
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /ˈkahni/ kahni
LA65.069 AP /moˈmoʔo/ momo’o
LA65.069 AP /mataˈsukwi/ matsukwi
LA65.069 AP /moˈmoʔo/ momo’o
LA65.069 AP /mataˈsukwi/ matsukwi
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /poʔoˈkadü/ po’okad
LA65.069 AP /poʔoˈkadü/ po’okad
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /poʔoˈkadü/ po’okad
LA65.069 AP /taʔniˈpüzi/ tanipüz 281
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /poʔoˈkadü/ po’okad
LA65.069 AP /pisüˈʔoːzi/ pish’oozi
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /püˈkeːdü/ pükeed
LA65.069 AP /ˈkahni/ kahni
LA65.069 AP /ʔeʔepiˈzijü/ e’epiziyu
LA65.069 AP /kaːˈdünü/ kaadünü
LA65.069 AP /kaːˈdümü/ kaadümü
LA65.069 AP /hoːˈkwidü/ hookwid
LA65.069 AP /hoːˈkwidü/ hookwid
LA65.069 AP /poʔoˈdümü/ po’odümü
LA65.069 AP /poʔoˈkweːdü/ po’okweed
LA65.069 AP /juˈɣwidü/ yugwid
LA65.069 AP /kahniˈjaːtü/ kahniyaat
LA65.069 AP /taˈvojo/ tavoyo
LA65.069 AP /pihaɣaˈmadü/ pihagamadü
LA65.069 AP /ʔüːˈvitü/ üüvitü
LA65.069 AP /hoːˈkwidü/ hookwid
LA65.069 AP /hoːˈkwidü/ hookwid
LA65.069 AP /paˈɣidü/ pagidü 282
LA65.069 AP /jaˈɣidü/ yadidü
LA65.069 AP /ʔeːˈpitsi/ eepich
LA65.069 AP /laˈmeːsa/ lameesa
LA65.069 AP /puˈɣuzi/ puguzi
LA65.069 AP /jühuːˈɣadü/ yühuugad
LA65.069 AP /ʔohoːˈɣadü/ ohoogad
LA65.069 AP /nahaˈjedü/ nahayedü
LA65.069 AP /kahniˈɣadü/ kahnigad
LA65.069 AP /kahniˈjaːtü/ kahniyaat
LA65.069 AP /puʔiˈjaːtü pu’iyaat
LA65.069 AP /ʔohoːˈwaːtü ohoowat
LA65.069 AP /kahniˈɣadü/ kahnigad
LA65.069 AP /woʔoˈravi/ wo’orav
LA65.069 AP /sünaˈʔasi/ süna’ash
LA65.069 AP /sünaˈʔasi/ süna’ash
LA65.069 AP /tühüjapijaˈviːna/ tühüyapiyaviina
LA65.069 AP /naˈɣavi/ nagavi
LA65.069 AP /ˈneːdü/ needü
LA65.069 AP /ʔuˈwarü/ uwaru
LA65.069 AP /kuˈtsapü/ kutsap
LA65.069 AP /kaˈʔadü/ ka’adü 283
LA65.069 AP /kaˈʔapü/ ka’apü
LA65.069 AP /ʔüˈʔapü/ ü’ap
LA65.069 AP /ʔüˈʔarü/ ü’arü
LA65.069 AP /tapitsiˈdiːka/ tapichidiika
LA65.069 AP /tapiˈtsidü/ tapichidü
LA65.069 AP /hiˈvidü/ hividü
LA65.069 AP /hiˈvidü/ hividü
LA65.069 AP /kwiˈdapü/ kwidap
LA65.069 AP /ˈhuːdü/ huudü
LA65.069 AP /ˈhuːpü/ huup
LA65.069 AP /siʔiˈɣweːdü/ shi’igweed
LA65.069 AP /siˈʔipü/ si’ip
LA65.069 AP /nüˈkadü/ nükadü
LA65.069 AP /nüˈkapü/ nükap
LA65.069 AP /koˈʔodü/ ko’odü
LA65.069 AP /kuˈkopi/ kukop
LA65.069 AP /kokoˈʔodü/ koko’odü
LA65.069 AP /poˈʔodü/ po’odü
LA65.069 AP /tuˈsudü/ tusudu
LA65.069 AP /ʔaˈriːna/ ariina
LA65.069 AP /manaˈɣidü/ managid 284
LA65.069 AP /ʔiˈsajdü/ isaydü
LA65.069 AP /tavasüˈtiːdü/ tavastiid
LA65.069 AP /kahnimanaˈɣidü/ kahnimanagid
LA65.069 AP /nasumaːˈɣadü/ nasumaagad
LA65.069 AP /hüˈʔadü/ hü’adü
LA65.069 AP /tübeːˈtiːdü/ tübeetiid
LA65.069 AP /tüˈbeːtü/ tübeet
LA65.069 AP /tübeːˈtiːdü/ tübeetiid
LA65.069 AP /tüˈbeːtü/ tübeet
LA65.069 AP /tüküpiˈwaːtü/ tüküpiwaat
LA65.069 AP /kaˈʔadü/ ka’adü
LA65.069 AP /kaʔaˈtüa/ ka’atüa
LA65.069 AP /kaˈrüdü/ karüdü
LA65.069 AP /karüˈnübü/ karünübü
LA65.069 AP /haˈvidü/ havidü
LA65.069 AP /haviˈtüa/ havitüa
LA65.069 AP /poˈʔodü/ po’odü
LA65.069 AP /poʔoˈnübü/ po’onübü
LA65.069 AP /jeˈʔedü/ ye’ed
LA65.069 AP /jeˈʔepü/ ye’ep
LA65.069 AP /nahajeˈʔedü/ nahaye’ed 285
LA65.069 AP /nahajeˈʔedü/ nahaye’ed
LA65.069 AP /taʔnipüziˈɣadü/ tanipüzigad
LA65.069 AP /pisüʔoːtsiˈɣadü/ pish’oochigad
LA65.069 AP /woˈkodü/ wokodü
LA65.069 AP /ʔiˈveːtü/ iveetü
LA65.069 AP /mutsuˈɣwidü/ mutsugwidü
LA65.069 AP /tuhuˈkwidü/ tuhukwidü
LA65.069 AP /tuhuˈkwidü/ tuhukwidü
LA65.069 AP /paʔatoˈɣodü/ pa’atogodü
LA65.069 AP /paʔatoˈɣodü/ pa’atogodü
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /seːˈɣidü/ seegid
LA65.069 AP /woˈkodü/ wokodü
LA65.069 AP /mutsuˈɣwidü/ mutsugwidü
LA65.069 AP /mutsuˈɣwidü/ mutsugwidü
LA65.069 AP /kaˈʔadü/ ka’adü
LA65.069 AP /kaˈʔapü/ ka’apü
LA65.069 AP /naːˈkeːdü/ naakeed
LA65.069 AP /püˈkeːdü/ pükeed
LA65.069 AP /pükeːˈdiːka/ pükeediika
LA65.069 AP /paˈɣidü/ pagidü 286
LA65.069 AP /wineːˈdiːka/ wineediika
LA65.069 AP /kaˈʔadü/ ka’adü
LA65.069 AP /kaˈʔadü/ ka’adü
LA65.069 AP /hoːˈkwidü/ hookwid
LA65.069 AP /tuɣunuˈkweːdü/ tugunukweed
LA65.069 AP /tuˈɣunu/ tugunu
LA65.069 AP /tahˈmana/ tahmana
LA65.069 AP /tahmanaˈkidü/ tahmanakid
LA65.069 AP /tuhuˈkwidü/ tuhukwidü
LA65.069 AP /tuhuiˈkweːdü/ tuhuikweed
LA65.069 AP /ʔaɣaˈkidü/ aaxkid
LA65.069 AP /ʔaɣaˈkweːdü/ aaxkweed
LA65.069 AP /wiˈjavü/ wiyav
LA65.069 AP /wijavüsüˈkweːdü/ wiyavüskweed
LA65.069 AP /manaˈɣidü/ managid
LA65.069 AP /maˈnidü/ manidü
LA65.069 AP /ˈpüːpü/ püüpü
LA65.069 AP /püːpüsiˈkweːdü/ püüpshikweed
LA65.069 AP /ʔaɣakiˈkweːdü/ aaxkikweed
LA65.069 AP /tuhuiˈkweːdü/ tuhuikweed
LA65.069 AP /tuhuiˈkweːdü/ tuhuikweed 287
LA65.069 AP /tuhukwiˈtiːna/ tuhukwitiina
LA65.069 AP /mahawaˈkweːdü/ mahawakweedü LA65.069 AP /kaːtiːˈdiːna/ kaatiidiina
LA65.069 AP /wijohwetiːˈdiːna/ wijohwetiidiina
LA65.069 AP /woˈkodü/ wokodü
LA65.069 AP /wokotiːˈniːka/ wokotiiniika
LA65.069 AP /hüˈʔütü/ hü’ütü
LA65.069 AP /ʔaɣaˈkidü/ aaxkid
LA65.069 AP /ʔaɣakitiːˈdiːka/ aaxkitidiika
LA65.069 AP /paʔatoˈɣodü/ pa’atogodü
LA65.069 AP /paʔatoɣotiːˈdiːka/ pa’atogotiidiika
LA65.069 AP /moɣowaˈriːdü/ mogowariidü
LA65.069 AP /moɣowariˈdiːka/ mogowaridiika
LA65.069 AP /moɣowaˈɣadü/ mogowagadü
LA65.069 AP /moɣowariˈdiːka/ mogowaridiika
LA65.069 AP /ˈpoʔo/ po’o
LA65.069 AP /sidoɣoˈʔopü/ shidogo’op
LA65.069 AP /sidoɣoʔoˈriːka/ shidogo’oriika
LA65.069 AP /niˈjaːmi/ niyaami
LA65.069 AP /nijaːriːˈdiːna/ niyariidiina
LA65.069 AP /wiˈjavü/ wiyav 288
LA65.069 AP /puhiˈɣidü/ puhigid
LA65.069 AP /puhiˈɣidü/ puhigid
LA65.069 AP /puhiɣitiːˈdiːka/ puhigitiidiika
LA65.069 AP /ʔuwatiːˈdiːka/ uwatiidiika
LA65.069 AP /paʔatoˈɣodü/ pa’atogodü
LA65.069 AP /paʔatoˈɣodü/ pa’atogodü
LA65.069 AP /pikaˈɣidü/ pikagid
LA65.069 AP /pikaˈɣidü/ pikagid
LA65.069 AP /paʔatoˈɣodü/ pa’atogodü
LA65.069 AP /taʔniˈpüzi/ tanipüz
LA65.069 AP /piˈsaːbü/ pishaabü
LA65.069 AP /ˈtiːpü/ tiipü
LA65.069 AP /ˈtiːpü/ tiipü
LA65.069 AP /ˈtübi/ tübi
LA65.069 AP /kwiˈdapü/ kwidap
LA65.069 AP /nahaˈjedü/ nahayed
LA65.069 AP /piˈsaːbü/ pishaabü
LA65.069 AP /piˈsaːbü/ pishaabü
LA65.069 AP /ʔaːʔabiɣiˈneːna/ aa’abigineena
LA65.069 AP /ʔaːʔabiɣiˈnaːma/ aa’abiginaama
LA65.069 AP /ʔaːweːɣiˈnaːmü/ aaweeginaamü 289
LA65.069 AP /ʔaːtakaˈpizi/ aatakapizh
LA65.069 AP /ʔaːpaɣiˈniːdü/ aapgaginiidü
LA65.069 AP /ʔüːˈvitü/ üüvitü
LA65.069 AP /ʔeːˈpizi/ eepizh
LA65.069 AP /ʔeːˈpitsi/ eepich
LA65.069 AP /ʔaːˈpiːnü/ aapiinü
LA65.069 AP /ʔüˈnüpü/ ünüpü
LA65.069 AP /ʔaːˈkaːdü/ aakaadü
LA65.069 AP /tuˈɣunu/ tugunu
LA65.069 AP /ʔaːtakaˈpizi/ aatakapizh
LA65.069 AP /siˈhwabü/ shihwabü
LA65.069 AP /ˈsoːmi/ soomi
LA65.069 AP /püraˈbümi/ Pürabüm
LA65.069 AP /ˈʔüːvi/ üüvi
LA65.069 AP /türüɣaˈviːka/ türügaviika
LA65.069 AP /kwiˈtsizi/ kwichizi
LA65.069 AP /ʔeˈɣubi/ egubi
LA65.069 AP /ʔeˈwutsi/ ewutsi
LA65.069 AP /ʔaɣaˈkidü/ aaxkid
LA65.070 AP /taʔniˈpüzi/ tanipüz
LA65.074 AP /nahniˈkweːdü/ nahnikweedü 290
LA65.074 AP /poʔoˈdümü/ po’odümü
LA65.074 AP /saˈmeːdü/ sameedü
LA65.075 AP /soʔoˈwaːtü/ so’owaat
LA65.075 AP /tukuˈʔaːmi/ tuku’aami
LA65.075 AP /toˈtsimi/ tochimi
LA65.075 AP /toˈtsimi/ tochimi
LA65.075 AP /ʔataˈvümi/ atavüm
LA65.075 AP /püraˈvümü/ püravüm
LA65.075 AP /kiːˈpuvü/ kiipuvü
LA65.075 AP /püraˈvümü/ püravüm
LA65.075 AP /moˈʔomi/ mo’omi
LA65.075 AP /maɣwinüˈbümi/ magwinübümi
LA65.075 AP /tasitoˈʔomi/ tashtomi
LA65.075 AP /naˈbini/ nabini
LA65.075 AP /ˈpüːpü/ püüpü
LA65.075 AP /jaɣaːˈdümi/ yagaadümi
LA65.075 AP /piˈhiːna/ pihiina
LA65.075 AP /saˈpiːna/ sapiina
LA65.075 AP /siˈʔipü/ si’ip
LA65.075 AP /kwaˈsiːna/ kwasiina
LA65.075 AP /saˈpiːna/ sapiina 291
LA65.075 AP /ʔaːˈpiːna/ aapiina
LA65.075 AP /püˈkeːdü/ pükeed
LA65.075 AP /jaˈɣidü/ yagidü
LA65.075 AP /naːˈkeːdü/ naakeed
LA65.075 AP /naːˈkeːdü/ naakeed
LA65.075 AP /weːˈɣipü/ weegipü
LA65.075 AP /navaˈkadü/ navakadü
LA65.075 AP /paˈtsaːtü/ patsaatü
LA65.075 AP /moʔoʔabiˈɣidü/ mo’o’abigidü
LA65.075 AP /tsawüruˈɣwidü/ tsawürugwidü
LA65.075 AP /nüɣaˈmidü/ nügamidü
LA65.075 AP /ʔabiˈɣidü/ abigidü
LA65.075 AP /weːɣiˈdümü/ weegidümü
LA65.075 AP /ʔabiɣiˈvistü/ abigivishtü
LA65.075 AP /ʔosoroˈjindü/ osorindü
LA65.075 AP /pihitsoɣoˈʔidü/ pihitsogo’idü
LA65.075 AP /haʔwiˈsidü/ ha’wisidü
LA65.075 AP /tsitsiˈjapü/ chichiyap
LA65.075 AP /ʔijavaˈɣadü/ iyavagad
LA65.075 AP /ʔijavaˈɣadü/ iyavagad
LA65.075 AP /ʔüijaˈɣadü/ üiyagad 292
LA65.075 AP /müːsüpiˈkweːdü/ müüsüpikweed
LA65.075 AP /jaˈɣidü/ yagid
LA65.075 AP /ʔabiˈɣipü/ abigip
LA65.075 AP /saˈmeːdü/ sameedü
LA65.075 AP /abiɣitiːdü/ abigitiid
LA65.075 AP /poˈʔodü/ po’odü
LA65.075 AP /poʔoˈnübü/ po’onübü
LA65.076 AP /naɣaˈdiːka/ nagadiika
LA65.076 AP /nasumaːkweːˈdiːka/ nasumaakweediika
LA65.076 AP /putsuɣuˈriːka/ putsuguriika
LA65.076 AP /ʔabiɣiˈdiːka/ abigidiika
LA65.076 AP /saˈmeːna/ sameena
LA65.076 AP /tüviˈsübi/ tüvisübi
LA65.076 AP /pünikeːˈdiːka/ pünikeediika
LA65.076 AP /kaːˈdümü/ kaadümü
LA65.076 AP /soˈʔodü/ so’odü
LA65.077 AP /ʔaˈwoho/ awoho
LA65.077 AP /totsaˈʔatü/ totsa’atü
LA65.077 AP /koʔokweːnˈdiːka/ ko’okweendiika
LA65.077 AP /koʔokweːnˈdiːka/ ko’okweendiika
LA65.077 AP /tonoˈdiːna/ tonodiina 293
LA65.077 AP /ʔasivoˈʔodü/ ashivo’odü
LA65.077 AP /ʔasivoʔoˈdiːka/ ashivo’odiika
LA65.077 AP /tonoˈnübü/ tononübü
LA65.077 AP /ʔˈaːza/ aaza
LA65.077 AP /moɣowaˈɣadü/ mogowagadü
LA65.077 AP /wüʔˈivü/ wü’ivü
LA65.077 AP /ˈtübi/ tübi
LA65.077 AP /puˈɣuzi/ puguzi
LA65.077 AP /ˈkeːvi/ keevi
LA65.077 AP /ˈtavi/ tavi
LA65.077 AP /tavikaːˈriːdü/ tavikaariidü
LA65.077 AP /ˈtavi/ tavi
LA65.077 AP /kaˈrüdü/ karüdü
LA65.077 AP /tuɣubijaːˈvidü/ tugubiyaavidü
LA65.077 AP /künaˈɣadü/ kunagadü
LA65.077 AP /ˈneːdü/ needü
LA65.077 AP /neːˈkidü/ neekidü
LA65.077 AP /toːˈhovü/ toohovü
LA65.077 AP /tazanoˈʔorü/ tazano’orü
LA65.078 AP /ˈtiːpü/ tiipü
LA65.078 AP /tahˈmana/ tahmana 294
LA65.078 AP /ˈtomo/ tomo
LA65.078 AP /ˈmüazi/ müazi
LA65.078 AP /taviweːniˈkweːpü/ taviweenikweep
LA65.078 AP /ˈʔüːvi/ üüvi
LA65.078 AP /ˈtavi/ tavi
LA65.078 AP /siˈmana/ shimana
LA65.078 AP /ˈʔüːvi/ üüvi
LA65.078 AP /piˈkaːju/ pikaayu
LA65.078 AP /taˈnasü/ tanasü
LA65.078 AP /juˈwaːtü/ yuwaat
LA65.078 AP /ˈsuːju/ suuyu
LA65.078 AP /waˈhaju/ wahayu
LA65.078 AP /aˈwatü/ awat
LA65.078 AP /hivoʔoˈpiːtsi/ hivo’opich
LA65.078 AP /tuˈɣubi/ tugubi
LA65.078 AP /ˈtoːvü/ toovü
LA65.078 AP /haviˈkweːdü/ havikweedü
LA65.078 AP /wüˈzünü/ wuzünü
LA65.078 AP /wünüˈkinü/ wünükinü
LA65.078 AP /poroˈnübü/ poronübü
LA65.078 AP /paɣiˈniːdü/ paginiidü 295
LA65.078 AP /mehepaɣiˈniːdü/ mehepaginiidü
LA65.078 AP /müziˈkweːdü/ muzikweedü
LA65.078 AP /pidüˈvaːdü/ pidüvaadü
LA65.078 AP /usaˈkweːdü/ uskweedü
LA65.078 AP /taʔniˈpüzi/ tanipüz
LA65.078 AP /tiːkweːˈvaːdü/ tiikweevaadü
LA65.078 AP /tiːˈkweːdü/ tiikweedü
LA65.078 AP /tsiˈpiːdü/ chipiidü
LA65.079 AP /tuˈɣunu/ tugunu
LA65.079 AP /tsaɣakweːˈdiːna/ tsagakweediina
LA65.079 AP /pünikeːˈdünü/ pünikeedünü
LA65.079 AP /püniküdüˈɣami/ püniküdügami
LA65.079 AP /weːˈdümü/ weedümü
LA65.079 AP /jawüˈdiːka/ yawüdiika
LA65.079 AP /paʔaˈvitü/ pa’avitü
LA65.079 AP /ˈtübi/ tübi
LA65.079 AP /taˈwiʔi/ tawi’i
LA65.079 AP /ʔivaˈʔana/ iva’ana
LA65.079 AP /ʔuˈweːnu/ uweenu
LA65.079 AP /ʔuˈweːnu/ uweenu
LA65.079 AP /jüˈhüvi/ yühüvi 296
LA65.079 AP /tiʔiˈdawi/ ti’idawi
LA65.079 AP /kuwapaʔaˈneːka/ kuwapa’aneeka
LA65.079 AP /türüɣaˈvika/ türügavika
LA65.079 AP /türüɣaˈvika/ türügavika
LA65.079 AP /türüɣaˈvika/ türügavika
LA65.079 AP /poʔoˈnübü/ po’onübü
LA65.079 AP /ʔaɣaˈkidü/ aaxkid
LA65.079 AP /aɣasaruˈɣwidü/ agasarugwidü
LA65.079 AP /aɣasaruˈɣwidü/ agasarugwidü
LA65.079 AP /puhiˈɣidü/ puhigid
LA65.079 AP /tosoˈkwidü/ tosokowidü
LA65.079 AP /ʔodoˈkwidü/ odokwidü
LA65.079 AP /ʔodoˈkwidü/ odokwidü
LA65.079 AP /puʔiˈɣadü/ pu’igadü
LA65.079 AP /tuhuˈkwidü/ tuhukwidü
LA65.079 AP /poinˈdeːdü/ poindeedü
LA65.079 AP /paʔatoɣoˈtiːna/ pa’atogotiina
LA65.079 AP /paˈʔadü/ pa’adü
LA65.079 AP /toveʔeˈpiːtsi/ tove’epich
LA65.079 AP /toveʔepiːtsiˈtiːna/ tove’epichtiina
LA65.079 AP /mutsuˈɣwidü/ mutsugwidü 297
LA65.079 AP /ʔohowaˈɣadü/ ohowagadü
LA65.079 AP /ʔaːsübiˈɣidü/ aasübigidü
LA65.079 AP /ʔaːsübiˈɣidü/ aasübigidü
LA65.079 AP /turaɣiˈtiːna/ turagitiina
LA65.079 AP /pikaˈɣidü/ pikagid
LA65.079 AP /poɣoˈzidü/ pogozidü
LA65.079 AP /poɣoˈzidü/ pogozidü
LA65.079 AP /poɣoˈzidü/ pogozidü
LA65.079 AP /totsivaʔˈaːtü/ totsiva’aatü
LA65.079 AP /juˈwaːtü/ yuwaat
LA65.079 AP /moɣowaˈɣadü/ mogowagadü
LA65.079 AP /tavasüˈkweːdü/ tavaskweedü
LA65.079 AP /tavasüˈkweːpü/ tavaskweepü
LA65.079 AP /taʔniˈpüzi/ tanipüz
LA65.079 AP /pisüˈʔoːzi/ pish’oozi
LA65.079 AP /tsaˈkaːvi/ tsakaavi
LA65.079 AP /ʔeːˈpizi/ eepizh
LA65.079 AP /ʔeːˈvizi/ eevizh
LA65.079 AP /neːzi/ neezh
LA65.079 AP /neːzidiˈkweːdü/ neezhdikweedü
LA65.079 AP /ʔaˈsaːzi/ asaazi 298
LA65.079 AP /ʔohoːˈwatü/ ohowaat
LA65.079 AP /kuhˈmaːtü/ kuhmatü
LA65.080 AP /peˈdiːna/ pediina
LA65.080 AP /paˈviːna/ paviina
LA65.080 AP /paˈviːna/ paviina
LA65.080 AP /paˈviːna/ paviina
LA65.080 AP /paˈziːna/ paziina
LA65.080 AP /peˈdiːna/ pediina
LA65.080 AP /kaɣuˈtsiːna/ kaguchiina
LA65.080 AP /kuhˈmani/ kuhmani
LA65.080 AP /pihwarüˈdiːna/ pihwarüdiina
LA65.081 AP /ˈtiːpü/ tiipü
LA65.081 AP /ˈtiːpü/ tiipü
LA65.081 AP /ˈkeːvi/ keevi
LA65.081 AP /keːˈpitsi/ keepich
LA65.081 AP /keːˈpitsi/ keepich
LA65.081 AP /taʔaˈwaːkü/ ta’awaakü
LA65.081 AP /ˈkeːvi/ keevi
LA65.081 AP /tünawaˈdiːka/ tünawadiika
LA65.081 AP /keːˈpitsi/ keepich
LA65.081 AP /ˈtübi/ tübi 299
LA65.081 AP /tüˈbitsi/ tübich
LA65.081 AP /sihˈwabü/ sihwabü
LA65.081 AP /wiˈjavü/ wiyav
LA65.081 AP /hoˈrodü/ horodü
LA65.081 AP /jüˈhürü/ yühürü
LA65.081 AP /sivoˈjaʔa/ shivoya’a
LA65.081 AP /koʔotoˈkadü/ ko’otokadü
LA65.081 AP /wiˈʔabi/ wi’abi
LA65.081 AP /kunapünikeːˈdiːka/ kunapünikeediika
LA65.081 AP /kuˈtsapi/ kutsap
LA65.081 AP /aɣaˈzidü/ agazidü
LA65.081 AP /aɣaˈzidü/ agazidü
LA65.081 AP /tüˈhüja/ tühüya
LA65.081 AP /ponoˈʔidü/ pono’idü
LA65.081 AP /kaˈʔapü/ ka’apü
LA65.081 AP /piˈhavi/ pihavi
LA65.081 AP /haːˈnizi/ haanizh
LA65.081 AP /noˈpavi/ nopavi
LA65.082 AP /karükweːˈvaːdü/ karükweevaadü
LA65.082 AP /ˈʔükü/ ükü
LA65.082 AP /sidoɣoˈkadü/ shidogokadü 300
LA65.082 AP /ˈkaɣi/ kagi
LA65.102 AP /siˈɣumi/ shigumi
LA65.102 AP /waˈʔapi/ wa’ap
LA65.102 AP /waˈʔapi/ wa’ap
LA65.102 AP /ʔiːˈkweːndü/ iikweendü
LA65.102 AP /ʔiːˈkweːndü/ iikweendü
LA65.102 AP /tsokoviˈʔina/ tsokovi’ina
LA65.102 AP /tsokoviˈʔivi/ tsokovi’ivi
LA65.102 AP /tsokoviˈʔina/ tsokovi’ina
LA65.102 AP /tsokoviˈʔivi/ tsokovi’ivi
LA65.102 AP /niːjoɣoˈdümü/ niiyogodümü
LA65.102 AP /niːjoɣoˈdümü/ niiyogodümü
tanipüz LA65.021 FC /taʔniˈpüzi/
LA65.021 FC /ˈtübi/ tübi
LA65.021 FC /ˈtübi/ tübi
LA65.021 FC /tüˈɣahni/ tügahni
LA65.021 FC /ˈtiːpü/ tiipü
LA65.021 FC /ˈtiːpü/ tiipü
LA65.021 FC /ˈtiːpü/ tiipü
LA65.021 FC /ˈtübi/ tübi 301
LA65.021 FC /ˈtiːpü/ tiipü
LA65.021 FC /ˈtiːpü/ tiipü
LA65.024 FC /ˈkahni/ kahni
LA65.024 FC /ˈkeːvi/ keevi
LA65.024 FC /ˈkuna/ kuna
LA65.024 FC /ˈkahni/ kahni
LA65.024 FC /ˈkuna/ kuna
LA65.024 FC /ˈtavi/ tavi
LA65.024 FC /ˈtaza/ taza
LA65.024 FC /ˈtiːpü/ tiipü
LA65.024 FC /ˈmüazi/ müazi
LA65.025 FC /woʔoˈravi/ wo’orav
LA65.025 FC /huˈvavi/ huvavi
LA65.025 FC /naˈvubü/ navubü
LA65.025 FC /hüaˈnübü/ hüanübü
LA65.025 FC /naˈvubü/ navubü
LA65.025 FC /huˈvavi/ huvavi
LA65.025 FC /hüaˈnübü/ hüanübü
LA65.025 FC /woʔoˈravi/ wo’orav
LA65.025 FC /kuˈtsapü/ kutsap
LA65.025 FC /kukopi/ kukop 302
LA65.025 FC /wiˈnapi/ winapi
LA65.025 FC /waˈʔapi/ wa’ap
LA65.025 FC /kwiˈhipü/ kwihip
LA65.025 FC /nüˈkapü/ nükap
LA65.025 FC /kuˈkopi/ kukop
LA65.025 FC /kwiˈhipü/ kwihip
LA65.025 FC /kuˈtsapü/ kutsap
LA65.025 FC /kuˈkopi/ kukop
LA65.025 FC /nüˈkapü/ nükap
LA65.025 FC /wiˈnapi/ winapi
LA65.025 FC /waˈʔapi/ wa’ap
LA65.025 FC /wiˈnapi/ winapi
LA65.025 FC /tuˈɣubi/ tugubi
LA65.025 FC /ʔeˈɣubi/ egubi
LA65.025 FC /taˈwabi/ tawabi
LA65.025 FC /ˈtaːbü/ taabü
LA65.025 FC /sihˈwabü/ sihwabü
LA65.025 FC /toˈtsivi/ tochivi
LA65.025 FC /taˈwabi/ tawabi
LA65.025 FC /ˈtaːbü/ taabü
LA65.025 FC /ˈtaːbü/ taabü 303
LA65.025 FC /taˈwabi/ tawabi
LA65.025 FC /sihˈwabü/ sihwabü
LA65.025 FC /wiˈhitsi/ wihich
LA65.025 FC /tuˈkutsi/ tukutsi
LA65.025 FC /muˈhutsi/ muhutsi
LA65.025 FC /moʔˈotsi/ mo’otsi
LA65.025 FC /tuˈkutsi/ tukutsi
LA65.025 FC /wiˈhitsi/ wihich
LA65.025 FC /tüˈɣahni/ tügahni
LA65.025 FC /peˈdüni/ pedüni
LA65.025 FC /tuˈwani/ tuwani
LA65.025 FC /muˈwani/ muwani
LA65.025 FC /piˈjani/ piyani
LA65.025 FC /tüˈɣahni/ tügahni
LA65.025 FC /küˈtsina/ küchina
LA65.025 FC /piˈjani/ piyani
LA65.025 FC /tüˈɣahni/ tügahni
LA65.025 FC /peˈdüni/ pedüni
LA65.025 FC /tuˈwani/ tuwani
LA65.025 FC /tüʔˈmadü/ tü’madü
LA65.025 FC /tühˈmarü/ tühmaru 304
LA65.025 FC /tahˈmana/ tahmana
LA65.025 FC /tühˈmarü/ tuhmarü
LA65.025 FC /tüʔˈmadü/ tü’madü
LA65.025 FC /tühˈmarü/ tühmarü
LA65.025 FC /tüʔˈmadü/ tü’madü
LA65.025 FC /tahˈmana/ tahmana
LA65.025 FC /ˈtavi/ tavi
LA65.025 FC /ˈtavi/ tavi
LA65.025 FC /ˈtavi/ tavi
LA65.025 FC /taˈwani/ tawani
LA65.025 FC /tahˈmana/ tahmana
LA65.025 FC /ˈtavi/ tavi
LA65.025 FC /tuˈwani/ tuwani
LA65.025 FC /tahˈmana/ tahmana
LA65.026 FC /ʔeˈɣubi/ egubi
LA65.026 FC /taˈwabi/ tawabi
LA65.026 FC /ˈtaːbü/ taabü
LA65.026 FC /kuˈtaːbü/ kutaabü
LA65.026 FC /sihˈwabü/ sihwabü
LA65.026 FC /küˈnabi/ künabi
LA65.026 FC /taːˈnüdü/ taanüdü 305
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /taːˈnüdü/ taanüdü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /taːˈnüdü/ taanüdü
LA65.026 FC /taːˈnüdü/ taanüdü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /ˈkahni/ kahni
LA65.026 FC /taːˈnüdü/ taanüdü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /taːˈnüdü/ taanüdü
LA65.026 FC /tüʔˈmarü/ tü’marü
LA65.026 FC /taːˈnüdü/ taanüdü
LA65.026 FC /tühˈmarü/ tühmarü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /tühˈmarü/ tühmarü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /tüʔˈmarü/ tü’marü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /taˈnadü/ tanadü
LA65.026 FC /tüʔˈmarü/ tü’marü
LA65.026 FC /naˈbini/ nabini 306
LA65.026 FC /naˈbivi/ nabivi
LA65.026 FC /naˈbimi/ nabimi
LA65.026 FC /naˈbini/ nabini
LA65.026 FC /naˈpovi/ napovi
LA65.026 FC /tüˈbivi/ tübivi
LA65.026 FC /tüˈbivi/ tübivi
LA65.026 FC /küˈsavi/ küsavi
LA65.026 FC /kuˈtsapü/ kutsap
LA65.026 FC /waˈtsuːju/ watsuuyu
LA65.026 FC /kuˈtsapü/ kutsap
LA65.026 FC /waˈtsuːju/ watsuuyu
LA65.026 FC /küˈsavi/ küsavi
LA65.026 FC /tuˈkutsi/ tukutsi
LA65.026 FC /nüˈkapü/ nükap
LA65.026 FC /nüˈkapü/ nükap
LA65.026 FC /wiziˈkizi/ wizikizi
LA65.026 FC /pisüˈʔoːzi/ pish’oozi
LA65.026 FC /pisüˈʔoːzi/ pish’oozi
LA65.026 FC /pisüˈʔoːzi/ pish’oozi
LA65.026 FC /pisüˈʔoːzi/ pish’oozi
LA65.027 FC /pawaˈhavü/ pawahavü 307
LA65.027 FC /taʔniˈpüzi/ tanipüz
LA65.027 FC /pajaˈnübü/ payanübü
LA65.027 FC /tasitoˈʔobi/ tasito’obi
LA65.027 FC /wiziˈkizi/ wizikizi
LA65.027 FC /neheˈkapi/ nehekapi
LA65.027 FC /naɣaˈvivi/ nagavivi
LA65.027 FC /naɣaˈvivi/ nagavivi
LA65.027 FC /naɣaˈvivi/ nagavivi
LA65.027 FC /naɣaˈvivi/ nagavivi
LA65.028 FC /moˈʔovi/ mo’ovi
LA65.028 FC /tahˈmana/ tahmana
LA65.028 FC /jüˈvana/ yüvana
LA65.028 FC /puˈɣuzi/ puguzi
LA65.028 FC /vaːˈkaʔa/ vaaka’a
LA65.028 FC /toˈroʔo/ toro’o
LA65.028 FC /tsivaʔaˈtoʔo/ chiva’ato’o
LA65.028 FC /noˈpavi/ nopavi
LA65.028 FC /kukopi/ kukop
LA65.028 FC /moˈkovü/ mokovü
LA65.028 FC /sünaˈʔavi/ süna’av
LA65.028 FC /wiˈnapi/ winapi 308
LA65.028 FC /winaˈhuwa/ winahuwa
LA65.028 FC /winaˈhuwa/ winahuwa
LA65.028 FC /huˈwazi/ huwazi
LA65.028 FC /kuˈtsapü/ kutsap
LA65.028 FC /ˈkuna/ kuna
LA65.028 FC /ˈkuna/ kuna
LA65.028 FC /kuˈtsapü/ kutsap
LA65.028 FC /ʔoˈhovü/ ohovü
LA65.028 FC /nüˈkapü/ nükap
LA65.028 FC /ʔaˈnisi/ anishi
LA65.028 FC /woʔˈnizi/ wo’nizhi
LA65.028 FC /hüaˈnübü/ hüanübü
LA65.028 FC /kwiˈhipi/ kwihip
LA65.028 FC /seːˈɣidü/ seegid
LA65.028 FC /taˈwabi/ tawabi
LA65.028 FC /wiˈhitsi/ wihichi
LA65.028 FC /paˈhazi/ pahazi
LA65.028 FC /ʔoˈsoro/ osoro
LA65.028 FC /ʔoˈsoro/ osoro
LA65.028 FC /hoˈjovi/ hoyovi
LA65.028 FC /paˈkwiːvi/ pakwiivi 309
LA65.028 FC /tuˈkutsi/ tukutsi
LA65.028 FC /waˈtsuːju/ watsuuyu
LA65.028 FC /mümüˈsuːju/ mümüsuuyu
LA65.028 FC /moˈʔoni/ mo’oni
LA65.028 FC /moˈʔomi/ mo’omi
LA65.028 FC /moˈʔovi/ mo’ovi
LA65.028 FC /toˈɣowa/ togowa
LA65.028 FC /paroˈɣowa/ parogowa
LA65.028 FC /puˈhavi/ puhavi
LA65.028 FC /woˈɣata/ wogata
LA65.028 FC /ʔeˈɣubi/ egubi
LA65.028 FC /naˈvubi/ navubi
LA65.028 FC /koˈvivi/ kovivi
LA65.028 FC /paˈjaːka/ payaaka
LA65.028 FC /tüˈɣahni/ tügahni
LA65.028 FC /wiˈhitsi/ wihici
LA65.028 FC /kwiˈtsizi/ kwichizhi
LA65.028 FC /huˈvavi/ huvavi
LA65.028 FC /tühˈmarü/ tühmarü
LA65.028 FC /kuˈtsapü/ kutsap
LA65.028 FC /joˈzidü/ yozidü 310
LA65.028 FC /jeˈʔedü/ ye’ed
LA65.028 FC /siʔiˈɣweːdü/ shi’igweed
LA65.028 FC /siˈʔidü/ shi’idü
LA65.028 FC /tüʔˈmadü/ tü’madü
LA65.028 FC /tüʔˈmadü/ tü’madü
LA65.028 FC /tüʔˈmadü/ tü’madü
LA65.028 FC /tsapuɣwiˈʔidü/ tsapuwi’idu
LA65.028 FC /kiːˈpuvü/ kiipuvü
LA65.028 FC /ˈsuːju/ suuyu
LA65.028 FC /ˈtiːpü/ tiipü
LA65.028 FC /ˈtavi/ tavi
LA65.028 FC /ˈsuːju/ suuyu
LA65.028 FC /tavi/ tavi
LA65.028 FC /ʔeːˈpizi/ eepizh
LA65.028 FC /neːˈzitsi neezhichi
LA65.028 FC /ʔeːˈpitsi/ eepich
LA65.028 FC /ʔeːˈpizi/ eepizh
LA65.028 FC /ʔeːˈpitsi/ eepich
LA65.028 FC /teːˈzitsi/ teezichi
LA65.028 FC /ˈneːzi/ neezh
LA65.028 FC /joˈzidü/ yozidü 311
LA65.028 FC /muːˈpizi/ muupizh
LA65.028 FC /pijaˈviːna/ piyaviina
LA65.028 FC /wijaˈrübü/ wiyarübü
LA65.028 FC /toːˈhovü/ toohovü
LA65.028 FC /pahaˈkoʔo/ pahako’o
LA65.028 FC /paˈmuhni/ pamuhni
LA65.028 FC /ˈkeːvi/ keevi
LA65.028 FC /keːˈpizi/ keepizh
LA65.028 FC /peˈzedü/ pezedü
LA65.028 FC /jaˈɣidü/ yagidü
LA65.028 FC /kuˈnabi/ kunabi
LA65.028 FC /puːˈtsiːvü/ puuchiivü
LA65.028 FC /ˈkeːvi/ keevi
LA65.028 FC /keːˈpizi/ keepizh
LA65.028 FC /paʔˈjütsi/ pa’yütsi
LA65.028 FC /ʔüʔaˈpüni/ ü’apüni
LA65.028 FC /nijaːˈɣadü/ niyaagadü
LA65.028 FC /ʔabiˈɣidü/ abigidü
LA65.028 FC /weːˈɣipü/ weegipü
LA65.028 FC /kijaˈrümü/ kiyarümü
LA65.028 FC /hüʔüjüˈtavi/ hü’üyütavi 312
LA65.028 FC /pajˈkweːdü/ paykweed
LA65.028 FC /joˈɣodü/ yogodü
LA65.028 FC /koˈvini/ kovini
LA65.028 FC /ʔaˈsaːzi/ asaazi
LA65.028 FC /nuˈvijo/ nuviyo
LA65.028 FC /wüˈʔivi/ wü’ivi
LA65.028 FC /keːvimaˈhaɣa/ keevimahaga
LA65.028 FC /naˈwabi/ nawabi
LA65.028 FC /tüˈbivi/ tübivi
LA65.028 FC /tüˈbimi/ tübimi
LA65.028 FC /tüˈbini/ tübini
LA65.028 FC /taːsuˈʔuvi/ taasu’uvi
LA65.028 FC /paˈvini/ pavini
LA65.028 FC /nüːwüˈʔami/ nüwü’ami
LA65.028 FC /ʔaˈjüːda/ ayüüda
LA65.028 FC /pohˈnija/ pohniya
LA65.028 FC /paˈɣüːzi/ pagüüzi
LA65.028 FC /paˈjaːka/ payaaka
LA65.028 FC /pajˈkweːdü/ paykweed
LA65.028 FC /pijaˈviːna/ piyaviina
LA65.028 FC /tüˈhüja/ tühüya 313
LA65.028 FC /poˈɣwitü/ pogwit
LA65.028 FC /wiˈʔabi/ wi’abi
LA65.028 FC /tsakaˈʔini/ tsaka’ini
LA65.028 FC /kosiˈkweːdü/ koshkweedü
LA65.028 FC /usaˈkweːdü/ uskweedü
LA65.028 FC /pajˈkweːdü/ paykweedü
LA65.028 FC /pajˈɣweːdü/ paygweedü
LA65.028 FC /pajˈkweːdü/ paykweedü
LA65.029 FC /ˈkahni/ kahni
LA65.029 FC /ˈkaːdü/ kaadü
LA65.029 FC /ˈkaːdü/ kaadü
LA65.029 FC /huvijaˈɣadü/ huviyagadü
LA65.029 FC /keːˈpizi/ keepizh
LA65.029 FC /keːˈpizi/ keepizh
LA65.029 FC /ʔaːsiˈbüzi aasibüzi
LA65.029 FC /ʔaːsiˈbüzi aasibüzi
LA65.029 FC /ʔaːsiˈbüzi aasibüzi
LA65.029 FC /kakaˈwutü/ kakawutü
LA65.029 FC /naːˈtüː/ naatüü
LA65.029 FC /ʔuːtsuˈkwasu/ uutsukwasu
LA65.029 FC /ʔohoˈwadü/ ohowadü 314
LA65.029 FC /noɣotseˈʔedü/ nogotse’edu
LA65.029 FC /hujuˈwaka/ huyuwaka
LA65.029 FC /naraʔüˈsaːɣa/ nara’üsaaga
LA65.029 FC /naraʔüˈsaːɣa/ nara’üsaaga
LA65.029 FC /naraʔüˈsaːɣa/ nara’üsaaga
LA65.029 FC /kotsoɣoˈʔodü/ kotsogo’odü
LA65.029 FC /kotsoɣoˈʔodü/ kotosogo’odü
LA65.029 FC /kaʔaˈvaːdü/ ka’avaadü
LA65.029 FC /hopaˈkidü/ hopakidü
LA65.029 FC /totsivaˈʔaːvü/ totsiva’aavü
LA65.029 FC /neʔeˈzitsi/ ne’ezhichi
LA65.029 FC /kiːˈpuvü/ kiipuvü
LA65.029 FC /pajaˈnübü/ payanübü
LA65.029 FC /naraʔüˈsaːɣa/ nara’üsaaga
LA65.029 FC /muvitoˈʔobü/ muvito’obü
LA65.029 FC /paˈɣüːzi/ pagüüzi
LA65.029 FC /taːsuˈʔuvi/ taasu’uvi
LA65.029 FC /pikotsaˈʔanzi/ pikotsa’anzi
LA65.029 FC /kaʔaˈvistü/ ka’avishtü
LA65.029 FC /waʔaˈdabü/ wa’adabü
LA65.029 FC /tüviziˈdaːra/ tüvizhidaara 315
LA65.029 FC /tsokopuˈʔisi tsokopu’ishi
LA65.029 FC /mataˈsukwi/ matsukwi
LA65.029 FC /mataˈsukwi/ matsukwi
LA65.029 FC /mataˈsukwi/ matsukwi
LA65.029 FC /ʔabiˈɣidü/ abigidü
LA65.029 FC /ʔabiˈɣidü/ abigidü
LA65.029 FC /kusüˈkwidü/ kuskwidü
LA65.029 FC /ʔosoroˈnidü/ osoronidü
LA65.029 FC /takaʔˈnadü/ taka’nadü
LA65.029 FC /kusüˈkwidü/ kuskwidü
LA65.029 FC /joˈzidü/ yozidü
LA65.029 FC /joziˈkweːdü/ yozikweedü
LA65.029 FC /jaˈɣidü/ yagidü
LA65.029 FC /moʔoˈɣadü/ mo’ogadü
LA65.029 FC /pariˈɣidü/ parigidü
LA65.029 FC /tsaniˈkwidü/ tsanikwidü
LA65.029 FC /mavajaˈkweːdü/ mavayakweedü
LA65.029 FC /joˈzidü/ yozidü
LA65.030 FC /ˈwija/ wiya
LA65.030 FC /kuˈtsapü/ kutsap
LA65.030 FC /ʔaˈsiva/ asiva 316
LA65.030 FC /ˈpüːpü/ püüpü
LA65.030 FC /ʔoˈhovü/ ohovü
LA65.030 FC /kuˈnavi/ kunavi
LA65.030 FC /hoʔiˈkidü/ ho’ikidü
LA65.030 FC /weʔeˈkweːdü/ we’ekweedü
LA65.030 FC /hiˈvidü/ hividü
LA65.030 FC /naɣaˈvivi/ nagavivi
LA65.030 FC /noˈpavi/ nopavi
LA65.030 FC /ˈnüʔü/ nü’ü
LA65.030 FC /ˈkuna/ kuna
LA65.030 FC /paˈɣüːzi/ pagüüzi
LA65.030 FC /joˈzidü/ yozidü
LA65.030 FC /muːˈpizi/ muupizhi
LA65.030 FC /puhiˈɣidü/ puhigid
LA65.030 FC /moˈʔovi/ mo’ovi
LA65.030 FC /ˈʔimi/ imi
LA65.030 FC /ˈnüʔü/ nü’ü
LA65.030 FC /poˈʔovi/ po’ovi
LA65.030 FC /taʔniˈpüzi/ tanipüz
LA65.030 FC /ʔüːˈvidü/ üüvidü
LA65.030 FC /tuˈɣunu/ tugunu 317
LA65.030 FC /muvitoˈʔobü/ muvito’obü
LA65.030 FC /ˈsuːju/ suuyu
LA65.030 FC /ʔuˈwarü/ uwaru
LA65.030 FC /püˈkeːdü/ pükeed
LA65.030 FC /ʔasiˈʔeːka/ ashieeka
LA65.030 FC /koʔotoˈkodü/ ko’otokodü
LA65.030 FC /puːˈtsiːvü/ puuchiivü
LA65.030 FC /ˈtavi/ tavi
LA65.030 FC /ˈʔimi/ imi
LA65.030 FC /ʔeˈɣubi/ egubi
LA65.030 FC /taˈwabi/ tawabi
LA65.030 FC /ˈpoʔo/ po’o
LA65.030 FC /ˈtami/ tami
LA65.030 FC /ponoˈʔidü/ pono’idu
LA65.030 FC /ˈmüazi/ müazi
LA65.030 FC /saˈpüvü/ sapüvü
LA65.030 FC /wiziˈkizi/ wizikizi
LA65.030 FC /tuhuˈkwidü/ tuhukwidü
LA65.030 FC /situˈʔidü/ situ’idü
LA65.030 FC /puˈɣuzi/ puguzi
LA65.030 FC /tavasüˈkweːdü/ tavasükweedü 318
LA65.030 FC /ˈtiːpü/ tiipü
LA65.030 FC /ˈtiːpü/ tiipü
LA65.030 FC /kaˈʔadü/ ka’adü
LA65.030 FC /wisiˈʔaːvü/ wishi’aavü
LA65.030 FC /naˈbivi/ nabivi
LA65.030 FC /tsoˈpivü/ tsopivü
LA65.030 FC /toˈtsivü/ totsivü
LA65.030 FC /naːˈkeːdü/ naakeed
LA65.030 FC /pühˈjüpü/ pühjüpü
LA65.030 FC /putsuˈɣudü/ putsugudü
LA65.030 FC /naˈɣeːka/ nageeka
LA65.030 FC /nüːˈwübi/ nüüwübi
LA65.030 FC /paʔatoˈɣodü/ pa’atogodü
LA65.030 FC /ʔaˈwatü/ awat
LA65.030 FC /ˈkeːvi/ keevi
LA65.030 FC /tüˈbivi/ tübivi
LA65.030 FC /kuˈravi/ kuravi
LA65.030 FC /juˈwaːtü/ yuwaat
LA65.030 FC /ʔaɣaˈkidü/ aaxkid
LA65.030 FC /sihˈwabü/ sihwabü
LA65.030 FC /ˈtübi/ tübi 319
LA65.030 FC /seːˈɣidü/ seegid
LA65.030 FC /ˈʔaːpü/ aapü
LA65.030 FC /ʔabiˈɣidü/ abigidü
LA65.030 FC /noɣotseˈʔedü/ nogotse’edü
LA65.030 FC /ˈtoːvü/ toovu
LA65.030 FC /ˈtoːvü/ toovü
LA65.030 FC /tasitoˈʔobi/ tasito’obi
LA65.030 FC /tasitoˈʔobi/ tasito’obi
LA65.030 FC /paːɣenuˈkwidü/ paagenukwidü
LA65.030 FC /kaˈrüdü/ karüdü
LA65.030 FC /paˈɣidü/ pagidü
LA65.030 FC /siˈʔidü/ si’idü
LA65.030 FC /wüˈnüdü/ wünüdü
LA65.030 FC /ˈhana/ hana
LA65.065 LG /taruˈʔidü/ taru’idü
LA65.065 LG /toːˈhovü/ toohovü
LA65.065 LG /taˈnimü/ tanimü
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /ˈkahni/ kahni
LA65.065 LG /kaˈrüdü/ karüdü 320
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /ˈtübi/ tübi
LA65.065 LG /taʔnipüˈzija/ tanipüziya
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /taʔnipüˈzija/ tanipüziya
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /hiˈvidü/ hividü
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /hoːˈkwidü/ hookwid
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /poʔoˈvaːdü/ po’ovaadü
LA65.065 LG /poʔovaːˈdümü/ po’ovaadümü
LA65.065 LG /kahniˈɣadü/ kahnigad
LA65.065 LG /kahniˈjaːtü/ kahniyaat
LA65.065 LG /taʔniˈpüzi/ tanipüz
LA65.065 LG /tüˈhüja/ tühüya
LA65.065 LG /küˈawe/ küawe
LA65.065 LG /ˈʔüːvi/ üüvi
LA65.065 LG /ʔivaˈʔana/ iva’ana
LA65.065 LG /ʔuˈweːnu/ uweenu 321
LA65.065 LG /kaʔaˈnübü/ ka’anübü
LA65.065 LG /kaʔaˈtüa/ ka’atüa
LA65.065 LG /taˈnimü/ tanimü
LA65.065 LG /ʔaɣaˈkidü/ aaxkid
LA65.065 LG /woʔoraˈvitsi/ wo’oravichi
LA65.065 LG /kaʔaˈnübü/ ka’anübü
LA65.065 LG /puɣuˈzitsi/ puguzichi
LA65.065 LG /kutuˈnüvü/ kutunüvü
LA65.065 LG /kutuˈnüvü/ kutunüvü
LA65.065 LG /ʔoˈhovü/ ohovü
LA65.065 LG /ʔoˈhovü/ ohovü
LA65.065 LG /ˈkahni/ kahni
LA65.065 LG /sünaˈʔasi/ süna’asi
LA65.065 LG /tübiɣahˈnizi/ tübigahnizhi
LA65.065 LG /kaːˈɣadü/ kaagadü
LA65.065 LG /kaʔaˈdümü/ ka’adümü
LA65.065 LG /kaʔaˈpüa/ ka’apüa
LA65.065 LG /ʔüˈʔapü/ ü’apü
LA65.065 LG /hiˈvidü/ hividü
LA65.065 LG /hiˈvidü/ hividü
LA65.065 LG /koˈʔorü/ ko’orü 322
LA65.065 LG /kaˈrüdü/ karüdü
LA65.065 LG /karüˈnübü/ karünübü
LA65.065 LG /haviˈtüa/ havitüa
LA65.065 LG /poʔoˈrü/ po’orü
LA65.065 LG /poʔoˈnübü/ po’onübü
LA65.065 LG /poʔoˈnübü/ po’onübü
LA65.066 LG /moˈmoʔo/ momo’o
LA65.066 LG /taʔniˈpüzi/ tanipüz
LA65.066 LG /pisüˈʔoːtsi/ pish’ootsi
LA65.066 LG /tuɣuˈkwidü/ tugukwidü
LA65.066 LG /paʔatoˈɣorü/ pa’atogorü
LA65.066 LG /kahniˈɣadü/ kahnigad
LA65.066 LG /ʔabiɣiˈdümü/ abigidümü
LA65.066 LG /ʔaːʔabiˈɣidü/ aa’abigidü
LA65.066 LG /püˈkeːdü/ pükeed
LA65.066 LG /muɣuwaˈrüdü/ muguwarüdü
LA65.066 LG /müːtsümuɣuwaˈrüdü/ müütsümuguwarüdü
LA65.066 LG /poˈʔorü/ po’orü
LA65.066 LG /kaˈʔarü/ ka’arü
LA65.066 LG /tuˈɣunu/ tugunu
LA65.066 LG /uːsutuɣunuˈkweːdü/ uusutugunukweedü 323
LA65.066 LG /tahˈmana/ tahmana
LA65.066 LG /kümiˈɣidü/ kümigidü
LA65.066 LG /mukujaˈɣadü/ mukuyagadü
LA65.066 LG /ʔuˈwarü/ uwaru
LA65.066 LG /püˈhüvü/ pühüvü
LA65.066 LG /ˈʔaːpü/ aapü
LA65.066 LG /soːˈkidü/ sookid
LA65.066 LG /maɣaʔaˈnidü/ maga’anidü
LA65.066 LG /maɣaˈhedü/ magahedü
LA65.066 LG /mapaʔniˈneːna/ mapa’nineena
LA65.066 LG /paˈtsaːtü/ patsaatü
LA65.066 LG /juˈwaːtü/ yuwaat
LA65.066 LG /tüˈhüja/ tühüya
LA65.067 LG /huˈwazi/ huwazi
LA65.067 LG /wiˈnapi/ winapi
LA65.067 LG /wiˈnapi/ winapi
LA65.067 LG /kuˈtsapü/ kutsap
LA65.067 LG /tuˈɣunu/ tugunu
LA65.067 LG /neheˈkapü/ nehekapü
LA65.067 LG /poˈɣwitü/ pogwit
LA65.067 LG /pisüʔˈoːzi/ pish’oozi 324
LA65.067 LG /pisüʔˈoːzi/ pish’oozi
LA65.067 LG /wiziˈkizi/ wizikizi
LA65.067 LG /taʔniˈpüzi/ tanipüz
LA65.067 LG /mataˈsivü/ matasivü
LA65.067 LG /ˈneːdü/ needü
LA65.067 LG /kwinuːˈrübü/ kwinuurübü
LA65.068 LG /sivoroˈnitsi/ shivoronichi
LA65.068 LG /sivoroˈnitsi/ shivoronichi
LA65.068 LG /ʔaˈnisi/ anishi
LA65.068 LG /ˈtiːpü/ tiipü
LA65.068 LG /sünaˈʔavi/ süna’av
LA65.068 LG /kohˈnotsi/ kohnotsi
LA65.068 LG /kohnoˈɣadü/ kohnogadü
LA65.068 LG /ʔüˈʔapü/ ü’apü
LA65.068 LG /wohoˈdübü/ wohodübü
LA65.068 LG /naruɣwaː'iˈkweːdü/ narugwaa’ikweedu
LA65.068 LG /peˈdüni/ pedüni
LA65.068 LG /tüˈhüja/ tühüya
LA65.068 LG /ʔaːtakaˈpizi/ aatakapizh
LA65.068 LG /tsiɣüˈpizi/ chigüpizhi
LA65.068 LG /puhiˈɣidü/ puhigid 325
LA65.068 LG /tsiɣüˈpizi/ chigüpizhi
LA65.068 LG /hutsiˈbija/ hutsibiya
LA65.068 LG /ˈmuhni/ muhni
LA65.068 LG /puˈʔivi/ pu’ivi
LA65.068 LG /sünaˈʔavi/ süna’av
LA65.068 LG /pijaˈviːna/ piyaviina
LA65.068 LG /ˈpüːpü/ püüpü
LA65.068 LG /ˈtiːpü/ tiipü
LA65.068 LG /kunaˈbüzi/ kunabüzi
LA65.068 LG /ˈneːdü/ needü
LA65.068 LG pohaɣadü pohagadü
LA65.068 LG /poʔoˈrü/ po’orü
LA65.068 LG /taʔniˈpüzi/ tanipüz
LA65.068 LG /muhˈnija/ muhniya
LA65.068 LG /ʔeːˈpizi/ eepizh
LA65.068 LG /tsaˈkaːvi/ tsakaavi
LA65.068 LG /tsaˈkaːvi/ tsakaavi
LA65.068 LG /moˈmoʔo/ momo’o
LA65.068 LG /ˈmuhni/ muhni
LA65.068 LG /tüˈhüja/ tühüya
LA65.068 LG /pisüʔˈoːzi/ pish’oozi 326
LA65.068 LG /taʔniˈpüzi/ tanipüz
LA65.068 LG /pisüʔoːˈzija/ pish’ooziya
LA65.068 LG /ˈkahni/ kahni
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /taʔniˈpüzi/ tanipüz
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /tüˈhüja/ tühüya
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /ˈtübi/ tübi
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /kaʔaˈtüa/ ka’atüa
LA65.068 LG /woʔoˈravi/ wo’orav
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /seːˈɣidü/ seegid
LA65.068 LG /piˈsaːbü/ pishaabü
LA65.068 LG /tavasüˈkweːpü/ tavasükweepü
LA65.068 LG /tavasüˈkweːpü/ tavasükweepü
LA65.068 LG /jühuːˈvistü/ yühuuvishtü
LA65.068 LG /ˈtübi/ tübi
LA65.068 LG /taʔniˈpüzi/ tanipüz 327
LA65.068 LG /poˈʔodü/ po’odü
LA65.068 LG /poˈʔodü/ po’odü
LA65.097 LG /ʔaˈwatü/ awat
LA65.097 LG /momoˈʔomü/ momo’omü
LA65.097 LG /ʔaˈwatü/ awat
LA65.097 LG /momoˈʔomü/ momo’omü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /taˈnimü/ tanimü
LA65.097 LG /ʔaˈwatü/ awat
LA65.097 LG /pisüʔˈoːzi/ pish’oozi
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔeʔepiˈziwü/ e’epizhiwü
LA65.097 LG /ʔaˈwatü/ awat
LA65.097 LG /ʔeviʔˈipitsu/ evi’ipichi
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔeːviˈziwü/ eevizhiwü
LA65.097 LG /ʔaˈwatü/ awat
LA65.097 LG /ʔüsaːˈziwü/ üsaaziwü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /mamaʔapüˈziwü/ mama’apüziwü
LA65.097 LG /ʔaˈwatü/ awat 328
LA65.097 LG /mamaʔapüˈziwü/ mama’apüziwü
LA65.097 LG /puˈɣuzi/ puguzi
LA65.097 LG /woʔoˈravi/ wo’orav
LA65.097 LG /wiziˈkizi/ wizikizi
LA65.097 LG /kakaˈwutü/ kakawutü
LA65.097 LG /kakaˈwutü/ kakawutü
LA65.097 LG /taˈvutsi/ tavutsi
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /taˈvutsi/ tavutsi
LA65.097 LG /sünaˈʔavi/ süna’av
LA65.097 LG /taʔniˈpüzi/ tanipüz
LA65.097 LG /taʔniˈpüzi/ tanipüz
LA65.097 LG /taʔniˈpüzi/ tanipüz
LA65.097 LG /tünijaˈkweːdü/ tüniyakweedü
LA65.097 LG /taʔniˈpüzi/ tanipüz
LA65.097 LG /tünijaˈkweːdü/ tüniyakweedü
LA65.097 LG /tünijaˈkweːdü/ tüniyakweedü
LA65.097 LG /taːsuˈʔuvi/ taasu’uvi
LA65.097 LG /taːsuˈʔuvi/ taasu’uvi
LA65.097 LG /taːsuˈʔuvi/ taasu’uvi
LA65.097 LG /taːsuˈʔuvi/ taasu’uvi 329
LA65.097 LG /tawaˈɣadü/ tawagadü
LA65.097 LG /tawaˈɣadü/ tawagadü
LA65.097 LG /tawaˈɣadü/ tawagadü
LA65.097 LG /tawaˈɣadü/ tawagadü
LA65.097 LG /tohoʔaˈɣadü/ toho’agadü
LA65.097 LG /tohoʔaˈɣadü/ toho’agadü
LA65.097 LG /tohoʔaˈɣadü/ toho’agadü
LA65.097 LG /tohoʔaˈɣadü/ toho’agadü
LA65.097 LG /tohoʔaˈɣadü/ toho’agadü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /ʔaˈwatü/ awatü
LA65.097 LG /tanaˈkweːdü/ tanakweedü
LA65.097 LG /tanaˈkweːdü/ tanakweedü
LA65.097 LG /taˈnarü/ tanarü
LA65.097 LG /tanakweːˈdiːka/ tanakweediika
LA65.097 LG /tuɣubaːˈɣadü/ tugubaagadü
LA65.097 LG /taˈvojo/ tavoyo 330
LA65.097 LG /taˈvojo/ tavoyo
LA65.097 LG /taˈvojo/ tavoyo
LA65.097 LG /tuɣubajaːˈɣadü/ tugubayaagadü
LA65.097 LG /tuɣubajaːvaˈʔadü/ tugubayaava’adü
LA65.097 LG /tuɣubajaːvaˈʔadü/ tugubaayava’adü
LA65.097 LG /tazanoˈʔorü/ tazano’orü
LA65.097 LG /tuːviˈɣidü/ tuuvigidü
LA65.097 LG /küˈnavi/ künavi
LA65.097 LG /küˈnavi/ künavi
LA65.097 LG /tazanoˈʔorü/ tazano’orü
LA65.097 LG /tazanoˈʔorü/ tazano’orü
LA65.097 LG /ˈtiːpü/ tiipü
LA65.097 LG /ˈtiːpü/ tiipü
LA65.097 LG /hüaˈnübü/ hüanübü
LA65.097 LG /ˈtüva/ tüva
LA65.097 LG /ˈtüva/ tüva
LA65.097 LG /ˈtüva/ tüva
LA65.097 LG /tüˈvapü/ tüvapü
LA65.097 LG /tüˈvapü/ tüvapü
LA65.097 LG /tüˈɣahni/ tügahni
LA65.097 LG /tüˈɣahni/ tügahni 331
LA65.097 LG /tüˈɣahni/ tügahni
LA65.097 LG /toˈɣowa/ togowa
LA65.097 LG /tsawaˈrübü/ tsawarübü
LA65.097 LG /tsawaˈrübü/ tsawarübü
LA65.097 LG /tsawaˈrübü/ tsawarübü
LA65.097 LG /tsawaˈrübü/ tsawarübü
LA65.097 LG /toˈtsivü/ totsivü
LA65.097 LG /toˈtsivü/ totsivü
LA65.097 LG /toˈtsivü/ totsivü
LA65.097 LG /toˈtsivü/ totsivü
LA65.097 LG /toˈtsivü/ totsivü
LA65.097 LG /toˈtsivü/ totsivü
LA65.097 LG /totsiˈvistü/ totsivishtü
LA65.097 LG /totsivaʔaːˈvistü/ totsiva’aavishtü
LA65.097 LG /nawüniˈnübü/ nawüninübü
LA65.097 LG /nawüniˈnübü/ nawüninübü
LA65.097 LG /nawüniˈnübü/ nawüninübü
LA65.097 LG /nawüniˈnübü/ nawüninübü
LA65.097 LG /totsivaʔaːˈɣadü/ totsiva’aagadü
LA65.097 LG /tsopivüʔaːˈɣadü/ tsopivü’aagadü
LA65.097 LG /totsivaʔaːˈɣadü/ totsiva’aagadü 332
LA65.097 LG /totsivaˈʔaːmi/ totsiva’aami
LA65.097 LG /totsivaˈʔaːmi/ totsiva’aami
LA65.097 LG /tutuˈpivü/ tutupivü
LA65.097 LG /tutuˈpivü/ tutupivü
LA65.097 LG /tutupivü/ tutupivü
LA65.097 LG /tuˈɣunu/ tugunu
LA65.097 LG /tuɣunuˈkweːdü/ tugunukweed
LA65.097 LG /wiʔˈabi/ wi’abi
LA65.097 LG /pahaˈrodü/ paharodü
LA65.097 LG /pahaˈrodü/ paharodü
LA65.097 LG /pijaˈviːna/ piyaviina
LA65.097 LG /tühüjapijaˈviːna/ tühüyapiyaviina
LA65.097 LG /tüˈhüja/ tühüya
LA65.097 LG /pijaˈviːna/ piyaviina
LA65.097 LG /wajahakaˈtübü/ wayahakatübü
LA65.097 LG /wajahakaˈtübü/ wayahakatübü
LA65.097 LG /wajahakaˈtübü/ wayahakatübü
LA65.097 LG /wajahakaˈtübü/ wayahakatübü
LA65.097 LG /wohoˈdübü/ wohodübü
LA65.097 LG /jüˈvibü/ yüvibü
LA65.097 LG /jüˈvibü/ yüvibü 333
LA65.097 LG /jüˈvibü/ yüvibü
LA65.098 LG /weːˈɣipü/ weegipü
LA65.098 LG /wijaˈrübü/ wiyarübü
LA65.098 LG /wijaˈrübü/ wiyarübü
LA65.098 LG /wijaˈrübü/ wiyarübü
LA65.098 LG /ˈwija/ wiya
LA65.098 LG /wiʔˈabi/ wi’abi
LA65.098 LG /ˈtiːpü/ tiipü
LA65.098 LG /wiˈjavi/ wiyavi
LA65.098 LG /wiˈjavi/ wiyavi
LA65.098 LG /wiʔˈabi/ wi’abi
LA65.098 LG /wiʔˈabi/ wi’abi
LA65.098 LG /ˈtiːpü/ tiipü
LA65.098 LG /waˈhaju/ wahayu
LA65.098 LG /ˈtiːpü/ tiipü
LA65.098 LG /wüʔˈivü/ wü’ivü
LA65.098 LG /wüʔˈivü/ wü’ivü
LA65.098 LG /woʔjeˈnübü/ wo’yenübü
LA65.098 LG /woʔoˈravi/ wo’orav
LA65.098 LG /patsaˈriːdü/ patsariidü
LA65.098 LG /patsaˈriːdü/ patsariidü 334
LA65.098 LG /koʔotoˈkorü/ ko’otokorü
LA65.098 LG /weːsaˈɣidü/ weesagidü
LA65.098 LG /weːsaˈɣidü/ weesagidü
LA65.098 LG /woˈkodü/ wokodü
LA65.098 LG /wiˈhitsi/ wihichi
LA65.098 LG /woˈkodü/ wokodü
LA65.098 LG /wiˈhitsi/ wihichi
LA65.098 LG /woˈkodü/ wokodü
LA65.098 LG /wiˈhitsi/ wihichi
LA65.098 LG /woˈkodü/ wokodü
LA65.098 LG /wiˈhitsi/ wihichi
LA65.098 LG /kwiˈtsizi/ kwichizi
LA65.098 LG /kwiˈtsizi/ kwichizi
LA65.098 LG /woɣodavaʔˈazi/ wogodava’azi
LA65.098 LG /woɣodavaʔˈazi/ wogodava’azi
LA65.098 LG /woɣodavaʔˈazi/ wogodava’azi
LA65.098 LG /wüˈnidü/ wünidü
LA65.098 LG /wüˈnizi/ wünizhi
LA65.098 LG /kahniwüˈnidü/ kahniwünidü
LA65.098 LG /kahniwüˈnidü/ kahniwünidü
LA65.098 LG /ʔüˈpiːdü/ üpiidü 335
LA65.098 LG /ʔüˈpiːdü/ üpiidü
LA65.098 LG /ˈpoʔo/ po’o
LA65.098 LG /wiziˈkizi/ wizikizi
LA65.098 LG /kahniˈɣadü/ kahnigad
LA65.098 LG /wiziˈkizi/ wizikizi
LA65.098 LG /kahniˈɣadü/ kahnigad
LA65.098 LG /paɣivaːˈdümü/ pagivaadümü
LA65.098 LG /paɣiniːˈvaːdü/ paginiivaadü
LA65.098 LG /tüˈhüja/ tühüya
LA65.098 LG /tsüpüɣiːˈnarü/ tsüpügiinarü
LA65.098 LG /naˈroʔo/ naro’o
LA65.098 LG /ʔatawaˈvaːdü/ atawavaadü
LA65.098 LG /müːsiˈpirü/ müüsipirü
LA65.098 LG /müːsiˈpirü/ müüsipirü
LA65.099 LG /koʔotoˈkorü/ ko’otokorü
LA65.099 LG /koʔotoˈkorü/ ko’otokorü
LA65.099 LG /ʔˈükü/ ükü
LA65.099 LG /ʔˈükü/ ükü
LA65.099 LG /ʔüküˈbüzi/ ükübüzi
LA65.099 LG /ˈhana/ hana
LA65.099 LG /warüˈmüpü/ warümüpü 336
LA65.099 LG /tüveːʔidaˈwitü/ tüvee’idawitü
LA65.099 LG /hüɣaˈriːka/ hügariika
LA65.099 LG /ʔüːvihüɣaˈriːka/ üüvihügariika
LA65.099 LG /ʔüːvihüɣaˈriːka/ üüvihügariika
LA65.099 LG /müˈriʔi/ müri’i
LA65.099 LG /müˈriʔi/ müri’i
LA65.099 LG /tasitoˈʔobi/ tasito’obi
LA65.099 LG /hüˈʔütü/ hü’ütü
LA65.099 LG /wüʔiˈvübü/ wü’ivübü
LA65.099 LG /muːtanaˈpizi/ muutanapizhi
LA65.099 LG /muːtanaˈpizi/ muutanapizhi
LA65.099 LG /muːtanaˈpizi/ muutanapizhi
LA65.099 LG /muːtanaˈpizi/ muutanapizhi
LA65.099 LG /nuˈkwidü/ nukwidü
LA65.099 LG /nuˈkwidü/ nukwidü
LA65.099 LG /wüˈnürü/ wünürü
LA65.099 LG /wüˈnürü/ wünürü
LA65.099 LG /wüˈnürü/ wünürü
LA65.099 LG /wüˈnürü/ wünürü
LA65.099 LG /nuˈkwidü/ nukwidü
LA65.099 LG /nuˈkwidü/ nukwidü 337
LA65.099 LG /nuˈkwidü/ nukwidü
LA65.099 LG /naˈvubü/ navubü
LA65.099 LG /naˈvubü/ navubü
LA65.099 LG /naˈvubü/ navubü
LA65.099 LG /navakaˈvaːdü/ mavakavaadü
LA65.099 LG /navakaˈvaːdü/ navakavaadü
LA65.099 LG /navaˈkarü/ navakarü
LA65.099 LG /navaˈkarü/ navakarü
LA65.099 LG /nuˈvavi/ nuvavi
LA65.099 LG /nuˈvavi/ nuvavi
LA65.099 LG /nuˈvavi/ nuvavi
LA65.099 LG /nuˈvavi/ nuvavi
LA65.099 LG /naraʔüsaːˈɣadü/ nara’üsaagadü
LA65.099 LG /naraʔüsaːˈɣadü/ nara’üsaagadü
LA65.099 LG /nohopüˈkedü/ nohopükedü
LA65.099 LG /nohopüˈkedü/ nohopükedü
LA65.099 LG /nohopüˈkedü/ nohopükedü
LA65.099 LG /niˈjaːni/ niyaani
LA65.099 LG /niˈjaːvi/ niyaavi
LA65.099 LG /niˈjaːvi/ niyaavi
LA65.099 LG /wiziˈkizi/ wizikizi 338
LA65.099 LG /joziˈkweːdü/ yozikweedü
LA65.099 LG /wiziˈkizi/ wizikizi
LA65.099 LG /kahnijaˈvümü/ kahniyavümü
LA65.099 LG /nohoˈvidü/ nohovidü
LA65.099 LG /nohoˈvidü/ nohovidü
LA65.099 LG /tapiˈtsidü/ tapichidü
LA65.099 LG /tapiˈtsidü/ tapichidü
LA65.099 LG /narapitsiˈkweːdü/ narapichikweedü
LA65.099 LG /narapitsiˈkweːdü/ narapichikweedü
LA65.099 LG /nijaˈnüdü/ niyanüdü
LA65.099 LG /nijaˈnüdü/ niyanüdü
LA65.099 LG /nijaˈnüdü/ niyanüdü
LA65.099 LG /nijaˈnüdü/ niyanüdü
LA65.099 LG /nawüniˈnübü/ nawüninübü
LA65.099 LG /wüˈnizi/ wünizhi
LA65.099 LG /nawüniˈnübü/ nawüninübü
LA65.099 LG /ponoʔˈidü/ pono’idü
LA65.099 LG /ʔabiɣiˈvistü/ abigivishtü
LA65.099 LG /ʔabiɣiˈvistü/ abigivishtü
LA65.099 LG /ʔuːsunukwiˈvaːdü/ uusunukwivaadü
LA65.099 LG /ʔuːsunukwiˈvaːdü/ uusunukwivaadü 339
LA65.099 LG /ʔuːsunukwiˈvaːdü/ uusunukwivaadü
LA65.099 LG /ʔuːsunukwiˈvaːdü/ uusunukwivaadü
LA65.099 LG /neʔeˈtiːdü/ ne’etiidü
LA65.099 LG /neʔeˈtiːdü/ ne’etiidü
LA65.099 LG /nanahaʔaˈrümü/ nanaha’arümü
LA65.099 LG /nanahaʔaˈrümü/ nanaha’arümü
LA65.099 LG /hiviˈnübü/ hivinübü
LA65.099 LG /hivinüˈbütsi/ hivinübutsi
LA65.099 LG /hivinüˈbütsi/ hivinübütsi
LA65.099 LG /pajaʔaˈnübü/ paya’anübü
LA65.099 LG /ˈnawi/ nawi
LA65.099 LG /pahoʔˈopi/ paho’opi
LA65.099 LG /pariˈɣidü/ parigidü
LA65.099 LG /pariˈɣidü/ parigidü
LA65.099 LG /pariˈɣidü/ parigidü
LA65.099 LG /pariˈɣidü/ parigidü
LA65.099 LG /pariˈɣidü/ parigidü
LA65.099 LG /pariˈɣidü/ parigidü
LA65.099 LG /püraˈvüni/ püravüni
LA65.099 LG /püraˈvüni/ püravüni
LA65.099 LG /paˈɣüːzi/ pagüüzi 340
LA65.099 LG /paˈɣüːzi/ pagüüzi
LA65.099 LG /paˈɣüːzi/ pagüüzi
LA65.099 LG /paˈɣüːzi/ pagüüzi
LA65.099 LG /paˈtsaːtü/ patsaatü
LA65.099 LG /paˈtsaːtü/ patsaatü
LA65.099 LG /kaˈrüdü/ karüdü
LA65.099 LG /ˈnüwü/ nüwü
LA65.099 LG /juɣwiˈdümü/ yugwidümü
LA65.099 LG /ˈnüːwü/ nüwü
LA65.099 LG /juɣwiˈdümü yugwidümü
LA65.099 LG /pihaɣaˈmadü/ pihagamadü
LA65.099 LG /pihavaniˈnübü/ pihavaninübü
LA65.099 LG /pipitahˈnirü/ pipitahnirü
LA65.099 LG /pipitahˈnirü/ pipitahnirü
LA65.099 LG /piˈsaːbü/ pishaabü
LA65.099 LG /ˈneːzi/ neezh
LA65.099 LG /piˈsaːbü/ pishaabü
LA65.099 LG /ˈneːzi/ neezh
LA65.099 LG /tuːrüˈɣeːdü/ tuurügeedü
LA65.099 LG /tuːrüˈɣeːdü/ tuurügeedü
LA65.099 LG /tuːrüˈɣeːdü/ tuurügeedü 341
LA65.099 LG /tuːrüˈɣeːdü/ tuurügeedü
LA65.099 LG /pitaˈnüsu/ pitanüsu
LA65.099 LG /pitaˈnüsu/ pitanüsu
LA65.099 LG /porotsiˈɣadü/ porotsigadü
LA65.099 LG /porotsiˈɣadü/ porotsigadü
LA65.099 LG /wiziˈkizi/ wizikizi
LA65.099 LG /wizikiˈzija/ wizikiziya
LA65.099 LG /hujuvaʔaˈdiːka/ huyuva’adiika
LA65.099 LG /ˈpoʔo/ po’o
LA65.099 LG /poʔonuˈkwidü/ po’onukwidü
LA65.099 LG /ˈpoʔonuˈkwidü/ po’onukwidü
LA65.099 LG /poʔopaːtoːˈrübü/ po’opaatoorübü
LA65.099 LG /poʔopaːtoːˈrübü po’opaatoorübü
LA65.099 LG /ˈnüʔü/ nü’ü
LA65.099 LG /nakaˈdiːka/ nakadiika
LA65.099 LG /poˈʔowa/ po’owa
LA65.099 LG /nukwiˈduwa/ nukwiduwa
LA65.099 LG /wiziˈkizi/ wizikizi
LA65.099 LG /koˈvija/ koviya
LA65.099 LG /mahaˈdiːka/ mahadiika
LA65.099 LG /koviˈjeːna/ koviyeena 342
LA65.099 LG /ˈnüʔü/ nü’ü
LA65.099 LG /puˈɣuzi/ puguzi
LA65.099 LG /pükeːˈdaːna/ pükeedaana
LA65.099 LG /puˈɣuzi/ puguzi
LA65.099 LG /puɣuˈzija/ puguziya
LA65.099 LG /puˈɣuzi/ puguzi
LA65.099 LG /pükeːdüˈrümü/ pükeedürümü
LA65.099 LG /puʔmitsaˈɣizi/ pu’michagizhi
LA65.099 LG /puʔmitsaɣiziˈvimü/ pu’michagizihvimü
LA65.099 LG /ʔaˈwatü/ awatü
LA65.099 LG /puˈɣuzi/ puguzi
LA65.099 LG /wiziˈkizi/ wizikizi
LA65.099 LG /wiziˈkizi/ wizikizi
LA65.099 LG /puːˈtsiːvü/ puuchiivü
LA65.099 LG /puːˈtsiːvü/ puuchiivü
LA65.099 LG /tüˈhüja/ tühüya
LA65.099 LG /tüˈhüja/ tühüya
LA65.099 LG /tüˈhüja/ tühüya
LA65.099 LG /putüsiːˈvümi/ putüsiivümi
LA65.099 LG /putüsiːˈvümi/ putüsiivümi
LA65.099 LG /pohaˈɣadü/ pohagadü 343
LA65.099 LG /pohaˈɣadü/ pohagadü
LA65.099 LG /pohaˈɣadü/ pohagadü
LA65.099 LG /puˈʔini/ pu’ini
LA65.099 LG /puˈʔini/ pu’ini
LA65.099 LG /puɣuˈzitsi puguzichi
LA65.099 LG /koʔotoˈkoro/ ko’otokoro
LA65.099 LG /saʔˈmapü/ sa’mapü
LA65.099 LG /saʔˈmapü/ sa’mapü
LA65.099 LG /juɣwiˈnübü/ yugwinübü
LA65.099 LG /juɣwiˈnübü/ yugwinübü
LA65.099 LG /juɣwiˈnübü/ yugwinübü
LA65.099 LG /seːˈɣidü/ seegid
LA65.099 LG /seːˈvübü/ seevübü
LA65.099 LG /ˈhiːbü/ hiibü
LA65.099 LG /hiːˈbüzi/ hiibüzi
LA65.099 LG /ˈhiːbü/ hiibü
LA65.099 LG /ˈhiːbü/ hiibü
LA65.099 LG /süɣüɣaˈmarü/ sügügamarü
LA65.099 LG /süɣüɣaˈmarü/ sügügamarü
LA65.099 LG /süɣüɣaˈmarü/ sügügamarü
LA65.099 LG /siɣoʔˈidü/ sigo’idü 344
LA65.099 LG /siɣoʔˈidü/ sigo’idü
LA65.099 LG /siɣoʔˈidü/ sigo’idü
LA65.099 LG /nümisiɣoʔiˈdümü/ nümisigo’idümü
LA65.099 LG /nümisiɣoʔiˈdümü/ nümisigo’idümü
LA65.099 LG /siɣoʔiˈdümü/ sigo’idümü
LA65.099 LG /siɣoʔiˈdümü/ sigo’idümü
LA65.099 LG /soːˈkidü/ sookid
LA65.099 LG /ˈsoːmi/ soomi
LA65.099 LG /ˈsoːmi/ soomi
LA65.099 LG /karüˈkadü/ karükadü
LA65.099 LG /pisüʔˈoːtsi/ pish’oochi
LA65.099 LG /sünaˈʔavi/ süna’av
LA65.099 LG /sünaˈʔavi/ süna’av
LA65.099 LG /sünaˈʔavi/ süna’av
LA65.099 LG /sünaˈʔavi/ süna’av
LA65.099 LG /situʔˈidü/ situ’idü
LA65.099 LG /situʔiˈkweːdü/ situ’ikweedü
LA65.099 LG /maˈhavü/ mahavü
LA65.099 LG /tsiːtoˈnorü/ chiitonorü
LA65.099 LG /noɣotseˈʔedü/ nogotse’edü
LA65.099 LG /tavivuːˈtsiːvü/ tavivuutsiivü 345
LA65.099 LG /tavivuːˈtsiːvü/ tavivuutsiivü
LA65.099 LG /tavivuːˈtsiːvü/ tavivuutsiivü
LA65.099 LG /kozoˈina/ kozoina
LA65.099 LG /kozoˈina/ kozoina
LA65.099 LG /taːˈrüna/ taarüna
LA65.099 LG /kozoˈina/ kozoina
LA65.099 LG /kozoˈina/ kozoina
LA65.099 LG /taːˈrüna/ taarüna
LA65.099 LG /tsaːruːˈina/ tsaaruuina
LA65.099 LG /ˈtübi/ tübi
LA65.099 LG /tüküpijaˈɣadü/ tüküpiyagadü
LA65.108 LG /poˈɣwitü/ pogwit
LA65.108 LG /paɣaˈzoːzi/ pagazoozi
LA65.108 LG /paɣaˈzoːzi/ pagazoozi
LA65.108 LG /paɣaˈzoːzi/ pagazoozi
LA65.108 LG /ʔaːwoˈɣobü/ aawogobü
LA65.108 LG /ʔaːwoˈɣobü/ aawagobü
LA65.109 LG /ʔaːˈpina/ aapina
LA65.109 LG /ʔaɣaˈkidü/ aaxkid
LA65.109 LG /ʔaɣaˈkidü/ aaxkid
LA65.109 LG /ʔaˈtsivi/ atsivi 346
LA65.109 LG /ʔaˈnivi/ anivi
LA65.109 LG /ʔataˈvümi/ atavümi
LA65.109 LG /ʔeˈwutsi/ ewutsi
LA65.109 LG /ʔeˈɣubi/ egubi
LA65.109 LG /ʔeˈɣubi/ egubi
LA65.109 LG /ʔoˈhovü/ ohovü
LA65.109 LG /ʔohowaˈɣadü/ ohowagadü
LA65.109 LG /ponoˈhorü/ ponohorü
LA65.109 LG /puːˈtsiːvü/ puuchiivü
LA65.109 LG /ʔaˈsaːzi/ asaazi
LA65.109 LG /tsiɣüˈpizi/ chigüpizhi
LA65.109 LG /tsiɣüˈpizi/ chigüpizhi
LA65.109 LG /huˈkubi/ hukubi
LA65.109 LG /huˈkubi/ hukubi
LA65.109 LG /sihˈwabü/ sihwabü
LA65.109 LG /sihˈwabü/ sihwabü
LA65.109 LG /sihˈwabü/ sihwabü
LA65.109 LG /toˈɣowa/ togowa
LA65.109 LG /hüˈvivi/ hüvivi
LA65.109 LG /ʔaraˈredü/ araredü
LA65.109 LG /kiˈjürü/ kiyürü 347
LA65.109 LG /ʔaraˈredü/ araredü
LA65.109 LG /kiˈjürü/ kiyürü
LA65.109 LG /joˈzidü/ yozidü
LA65.109 LG /jühuˈɣadü/ yühugadü
LA65.109 LG /jüˈhuːvü/ yühuuvü
LA65.109 LG /jüˈvana/ yüvana
LA65.109 LG /kwaˈsivi/ kwasivi
LA65.109 LG /kwaˈsivi/ kwasivi
LA65.109 LG /ˈkaːzi/ kaazi
LA65.109 LG /kuˈravi/ kuravi
LA65.109 LG /kwiˈhipi/ kwihip
LA65.109 LG /kwiˈhipi/ kwihip
LA65.109 LG /ˈkuna/ kuna
LA65.109 LG /koˈvivi/ kovivi
LA65.109 LG /kaˈrüdü/ karüdü
LA65.109 LG /kaˈʔadü/ ka’adü
LA65.109 LG /kotsoɣoʔˈadü/ kotsogo’adü
LA65.109 LG /kotsoɣoʔˈadü/ kotsogo’adü
LA65.109 LG /kotsoɣoʔˈadü/ kotsogo’adü
LA65.109 LG /ʔatakoʔaˈnübü/ atako’anübü
LA65.109 LG /ʔatakoʔaˈnübü/ atako’anübü 348
LA65.109 LG /ʔatakoʔaˈnübü/ atako’anübü
LA65.109 LG /ʔatakoʔanüˈbüni/ atako’anübüni
LA65.109 LG /kiːˈpuni/ kiipuni
LA65.109 LG /kopakaˈtiːdü/ kopakatiid
LA65.109 LG /joˈzidü/ yozidü
LA65.109 LG /muːˈpizi/ muupizhi
LA65.109 LG /muˈhutsi/ muhutsi
LA65.109 LG /müˈjütsi/ müyütsi
LA65.109 LG /mataˈsukwi/ matsukwi
LA65.109 LG /mataˈsukwi/ matsukwi
LA65.109 LG /muˈwani/ muwani
LA65.109 LG /ʔivaʔˈana/ iva’ana
LA65.109 LG /naɣaˈvivi/ nagavivi
LA65.109 LG /navaˈhaju/ navhayu
LA65.109 LG /noːˈsidü/ noosidü
LA65.109 LG /ˈnüːwü/ nüwü
LA65.109 LG /nüˈkarü/ nükarü
LA65.109 LG /nükaˈtüa/ nükatüa
LA65.109 LG /nuˈvavi/ nuvavi
LA65.109 LG /nuˈvavi/ nuvavi
LA65.109 LG /nuˈvavi/ nuvavi 349
LA65.109 LG /neʔˈedü/ ne’edü
LA65.109 LG /nuˈvavi/ nuvavi
LA65.109 LG /nuˈvavi/ nuvavi
LA65.109 LG /nawaˈɣarü/ nawagarü
LA65.109 LG /paˈɣüːzi/ pagüüzi
LA65.109 LG /poroˈɣowa/ porogowa
LA65.109 LG /paˈvini/ pavini
LA65.109 LG /pawaˈhavü/ pawahavü
LA65.109 LG /ˈpüːpü/ püüpü
LA65.109 LG /paˈkadü/ pakadü
LA65.109 LG /peˈdüni/ pedüni
LA65.109 LG /piˈjani/ piyani
LA65.109 LG /piˈjani/ piyani
LA65.109 LG /wijabiˈkini/ wiyabikini
LA65.109 LG /wijabiˈkini/ wiyabikini
LA65.109 LG /pohˈnija/ pohniya
LA65.109 LG /pohˈnija/ pohniya
LA65.109 LG /puˈʔivi/ pu’ivi
LA65.109 LG /pohˈnija/ pohniya
LA65.109 LG /pohˈnija/ pohniya
LA65.109 LG /ʔaːsiˈbüzi/ aasibüzi 350
LA65.109 LG /puhiˈɣidü/ puhigid
LA65.109 LG /puˈɣuzi/ puguzi
LA65.109 LG /puʔmitsaˈɣizi/ pu’michagizhi
LA65.109 LG /puʔmitsaˈɣizi/ pu’michagizhi
LA65.109 LG /püraˈvüni/ püravüni
LA65.109 LG /püˈhübü/ pühübü
LA65.109 LG /siɣoʔˈidü/ sigo’idü
LA65.109 LG /siɣoʔˈidü/ sigo’idü
LA65.109 LG /ˈsoːvi/ soovi
LA65.109 LG /soːˈkidü/ sookid
LA65.109 LG /tahˈmana/ tahmana
LA65.109 LG /tahˈmana/ tahmana
LA65.109 LG /ˈtaza/ taza
LA65.109 LG /ˈtaza/ taza
LA65.109 LG /taˈnavü/ tanavü
LA65.109 LG /taˈnani/ tanani
LA65.109 LG /tasitoˈʔobi/ tasito’obi
LA65.109 LG /ˈtavi/ tavi
LA65.109 LG /tasitoʔˈobi/ tasito’obi
LA65.109 LG /taˈwabi/ tawabi
LA65.109 LG /taˈwabi/ tawabi 351
LA65.109 LG /tiʔidaˈwitü/ ti’idawitü
LA65.109 LG /ˈtomo/ tomo
LA65.109 LG /seːˈɣidü/ seegid
LA65.109 LG /ˈtübi/ tübi
LA65.109 LG /tuhuˈkwidü/ tuhukwidü
LA65.109 LG /tuˈɣunu/ tugunu
LA65.109 LG /tuˈɣunu/ tugunu
LA65.109 LG /ˈtavi/ tavi
LA65.109 LG /ʔeːˈpitsi/ eepich
LA65.109 LG /tsakaːˈviʔi/ tsakaavi’i
LA65.109 LG /tuˈwani/ tuwani
LA65.109 LG /tüˈbimi/ tübimi
LA65.109 LG /tüˈbimi/ tübimi
LA65.109 LG /ˈtübi/ tübi
LA65.109 LG /ˈtiːpü/ tiipü
LA65.109 LG /waˈtsuːju/ watsuuyu
LA65.109 LG /woˈɣata/ wogata
LA65.109 LG /wiˈhitsi/ wihichi
LA65.109 LG /woˈzija/ woziya
LA65.109 LG /kutsaˈkidü/ kutsakidü
LA65.109 LG /kaʔˈadü/ ka’adü 352
LA65.109 LG /mataˈsukwi/ matsukwi
LA65.109 LG /huˈwazi/ huwazi
LA65.109 LG /puˈɣuzi/ puguzi
LA65.109 LG /muˈwani/ muwani
LA65.109 LG /namiʔˈini/ nami’ini
LA65.109 LG /namiʔˈini/ nami’ini
LA65.109 LG /paˈvini/ pavini
LA65.109 LG /tsakaʔˈini/ tsaka’ini
LA65.109 LG /ʔamiˈɣoni/ amigoni
LA65.109 LG /tsijaˈkani/ chiyakani
LA65.109 LG /seːˈɣidü/ seegid
LA65.109 LG /taʔniˈpüzi/ tanipüz
LA65.109 LG /ˈnüwü/ nüwü
LA65.109 LG /ʔaˈsa:zi/ asaazi
LA65.109 LG /maʔaˈpüzi/ ma’apüzi
LA65.109 LG /toˈtsimi/ tochimi
LA65.109 LG /totsivaʔˈaːnü/ totsiva’aanü
LA65.109 LG /mutaˈkami/ mutakami
LA65.109 LG /soviˈvümü/ sovivümü
LA65.109 LG /ʔataˈvümi/ atavümi
LA65.109 LG /atakoʔanaˈbümi/ atako’anabümi 353
LA65.109 LG /naɣaˈvini/ nagavini
LA65.109 LG /puˈʔimi/ pu’imi
LA65.109 LG /muvitoʔˈomi/ muvito’omi
LA65.109 LG /tüˈbimi/ tübimi
LA65.109 LG /taˈwami/ tawami
LA65.109 LG /tüˈbimi/ tübimi
LA65.109 LG /mizoˈɣadü/ mizhogadü
LA65.109 LG /ʔeˈɣubi/ egubi
LA65.109 LG /kuˈrani/ kurani
LA65.109 LG /tsoːˈvümi/ tsoovümi
LA65.109 LG /pisipoʔˈomi/ pishipo’omi
LA65.109 LG /huvakaweʔˈedü/ huvakawe’edü
LA65.109 LG /opijarüʔˈodü/ opiyarü’odü
LA65.109 LG /jühuːˈɣadü/ yühuugad
LA65.109 LG /ʔasiʔˈami/ asi’ami
LA65.109 LG /wisiʔaːˈɣadü/ wishi’aagadü
LA65.109 LG /noˈpavi/ nopavi
LA65.109 LG /püˈhüvü/ pühüvü
LA65.109 LG /tasitoˈʔobi/ tasito’obi
LA65.109 LG /tuhuˈkwidü/ tuhukwidü
LA65.109 LG /puhiˈɣidü/ puhigid 354
LA65.109 LG /ʔaɣaˈkidü/ aaxkid
LA65.109 LG /seːˈɣidü/ seegid
LA65.109 LG /ˈpoʔo/ po’o
LA65.109 LG /tazanoʔˈorü/ tazano’orü
LA65.109 LG /aɣazinʔˈorü/ agazin’orü
LA65.109 LG /paːɣenuˈkwidü/ paagenukwidü
LA65.109 LG /ʔodoˈkwidü/ odokwidü
LA65.109 LG /ʔodoˈkwidü/ odokwidü
LA65.110 LG /tukuːˈmüːtsi/ tukuumüüts
LA65.110 LG /tuˈkutsi/ tukutsi
LA65.110 LG /tüˈvizi/ tüvizhi
LA65.110 LG /sünaˈʔavi/ süna’av
LA65.110 LG /puˈɣuzi/ puguzi
LA65.110 LG /pohˈnija/ pohnia
LA65.110 LG /taˈvutsi/ tavutsi
LA65.110 LG /tavuˈtsitsi/ tavutsichi
LA65.110 LG /ˈkamü/ kamü
LA65.110 LG /ʔeˈwutsi/ ewutsi
LA65.110 LG /woˈzija/ woziya
LA65.110 LG /ˈkaːzi/ kaazi
LA65.110 LG /puʔmitsaˈɣizi/ pu’michagizhi 355
LA65.110 LG /müˈjütsi/ müyütsi
LA65.110 LG /paːtsaʔˈaːzi/ paatsa’aazi
LA65.110 LG /paˈɣüːzi/ pagüüzi
LA65.110 LG /paˈɣüːzi/ pagüüzi
LA65.110 LG /paˈɣüːzi/ pagüüzi
LA65.110 LG /woˈɣata/ wogata
LA65.110 LG /tsiɣüˈpizi/ chigupizhi
LA65.110 LG /ˈʔaja/ aya
LA65.110 LG /haːˈnizi/ haanizhi
LA65.110 LG /ʔaːtakaˈpizi/ aatakapizh
LA65.110 LG /huˈkubi/ hukubi
LA65.110 LG /muˈhutsi/ muhutsi
LA65.110 LG /ʔataˈkazi/ atakazi
LA65.110 LG /tsiˈɣazi/ tsigazi
LA65.110 LG /hüˈʔütü/ hü’ütü
LA65.110 LG /ʔiˈveːtü/ iveetü
LA65.110 LG /ʔaˈtaːbü/ ataabü
LA65.110 LG /ʔüːveʔeˈpiːtsi/ üüve’epich
LA65.110 LG /paʔatoˈɣodü/ pa’atogodü
LA65.110 LG /toveʔeˈpiːtsi/ tove’epich
LA65.110 LG /taruˈʔidü/ taru’idü 356
LA65.110 LG /situˈʔidü/ situ’idü
LA65.110 LG /neheˈkapi/ nehekapi
LA65.110 LG /tüˈbeːtü/ tübeetü
LA65.110 LG /tavasüˈkweːpü/ tavasükweepü
LA65.110 LG /tavasüˈkweːpü/ tavasükweepü
LA65.110 LG /navaˈhaju/ navahayu
LA65.110 LG /pasoːˈzidü/ pasoozidü
LA65.110 LG /paʔatoˈɣorü/ pa’atogorü
LA65.110 LG /tüɣüjeʔˈerü/ tügüye’erü
LA65.110 LG /nüˈkarü/ nükarü
LA65.110 LG /püˈkeːdü/ pükeed
LA65.110 LG /noːˈsidü/ noosidü
LA65.110 LG /pajaˈʔarü/ paya’arü
LA65.110 LG /türüˈmaːrü/ türümaarü
LA65.110 LG /türümaːˈkweːdü/ türümaakweedü
LA65.110 LG /joˈzidü/ yozidü
LA65.110 LG /kaˈrüdü/ karüdü
LA65.110 LG /ʔüˈpiːdü/ üpiidü
LA65.110 LG /ponoˈhorü/ ponohorü
LA65.110 LG /wüˈnürü/ wünürü
LA65.110 LG /paˈɣidü/ pagidü 357
LA65.110 LG /kuˈkopi/ kukopi
LA65.110 LG /ʔabiˈɣidü/ abigidü
LA65.110 LG /tüniˈjadü/ tüniyadü
LA65.110 LG /muɣuwaˈrüdü/ muguwarüdü
LA65.110 LG /taˈvidü/ tavidü
LA65.110 LG /taˈvidü/ tavidü
LA65.110 LG /nijaˈnüdü/ niyanüdü
LA65.110 LG /ʔiniˈɣadü/ inigadü
LA65.110 LG /pisüʔoːrüˈvaːdü/ pish’oorüvaadü
LA65.110 LG /paɣiˈkweːdü/ pagikweedü
LA65.110 LG /ʔusaˈkwidü/ uskwidü
LA65.110 LG /ʔusaˈkweːdü/ uskweedü
LA65.110 LG /ʔohowaʔabiˈɣidü/ ohowa’abigidü
LA65.110 LG /joziˈkidü/ yozikidü
LA65.110 LG /kiˈjarü/ kiyarü
LA65.110 LG /ʔusaˈkweːdü/ usakweedü
LA65.110 LG /kwitarüniˈjadü/ kwitaruniyadü
LA65.110 LG /niˈjaːmi/ niyaami
LA65.110 LG /soːˈkwidü/ sookidü
LA65.110 LG /nuˈkwidü/ nukwidü
LA65.110 LG /paˈkarü/ pakarü 358
LA65.110 LG /hoːˈkwidü/ hookwid
LA65.110 LG /kuˈkwidü/ kukwidü
LA65.110 LG /tüˈhüja/ tühüya
LA65.110 LG /kaʔaˈvaːna/ ka’avaana
LA65.110 LG /hiviˈvaːna/ hivivaana
LA65.110 LG /kopakiˈtiːdü/ kopakitiidü
LA65.110 LG /noɣotseˈʔedü/ nogotse’edü
LA65.110 LG /nawaˈɣarü/ nawagarü
LA65.110 LG /jaˈɣidü/ yagidü
LA65.110 LG /koʔokweːˈnüdü/ ko’okweenüdü
LA65.110 LG /nukwiˈtiːdü/ nukwitiidü
LA65.110 LG /pihaɣaˈmadü/ pihagamadü
LA65.110 LG /ʔowaɣaˈmadü/ owagamadü
LA65.110 LG /süɣüɣaˈmadü/ sügügamadü
LA65.110 LG /ʔoˈwavi/ owavi
LA65.110 LG /piˈhavi/ pihavi
LA65.110 LG /wiˈʔabi/ wi’abi
LA65.110 LG /wiˈʔabi/ wi’abi
LA65.120 LG /ʔaɣaˈkidü/ aaxkid
LA65.120 LG /seːˈɣidü/ seegid
LA65.120 LG /tuhuˈkwidü/ tuhukwidü 359
LA65.120 LG /sanaʔoːˈtsozi/ sana’ootsozi
LA65.120 LG /sanaʔoːˈtsozi/ sana’ootsozi
LA65.120 LG /puːˈtsiːvü/ puutsiivü
LA65.120 LG /puːˈtsiːvü/ puutsiivü
LA65.120 LG /ˈmüazi/ müazi
LA65.120 LG /ˈtavi/ tavi
LA65.120 LG /ˈtavi/ tavi
LA65.120 LG /juˈwaːtü/ yuwaat
LA65.120 LG /ˈtiːpü/ tiipü
LA65.120 LG /puˈɣuzi/ puguzi
LA65.120 LG /juˈwaːtü/ yuwaat
LA65.120 LG /weːɣiˈdümü/ weegidümü
LA65.074 RB /ˈkarü/ karü
LA65.075 RB /toˈtsimi/ totsimi
LA65.075 RB /nakaˈvimi/ nakavimi
LA65.075 RB /tüˈbimi/ tübimi
LA65.075 RB /ʔeɣubi/ egubi
LA65.075 RB /püraˈvümü/ püravümü
LA65.075 RB /püraˈvümü/ püravümü
LA65.075 RB /tasitoˈʔobi/ tasito’obi
LA65.075 RB /naˈbimi/ nabimi 360
LA65.075 RB /siˈʔipü/ si’ip
LA65.075 RB /püˈkeːna/ pükeena
LA65.075 RB /naːkeːˈdiːka/ naakeediika
LA65.075 RB /soːˈkidü/ sookid
LA65.075 RB /tawaˈbümi/ tawabümi
LA65.075 RB /nüɣaˈmüdü/ nügamüdü
LA65.075 RB /nüɣaˈmüdü/ nügamüdü
LA65.075 RB /weːɣiˈdümü/ weegidümü
LA65.075 RB /ʔabiɣiˈjaːtü/ abigiyaat
LA65.075 RB /ʔabiˈɣidü/ abigidü
LA65.075 RB /kusükwiˈɣadü/ kusükwigadü
LA65.075 RB /kiˈjüdü/ kiyüdü
LA65.075 RB /kiˈjüdü/ kiyüdü
LA65.075 RB /kiˈjüdü/ kiyüdü
LA65.075 RB /taːˈnüdü/ taanüdü
LA65.075 RB /pihitsoɣoˈʔidü/ pihitsogo’idü
LA65.075 RB /peːˈdiːna/ peediina
LA65.075 RB /ʔüijaˈɣadü/ üiyagadü
LA65.075 RB /ʔüijaˈɣadü/ üiyagadü
LA65.075 RB /pisaweːˈdiːna/ pisaweediina
LA65.075 RB /tüɣüjeˈɣadü/ tügüyegadü 361
LA65.075 RB /nünüjeˈʔedü/ nünüye’edü
LA65.075 RB /ʔabiɣitiːˈnübü/ abigitiinübü
LA65.075 RB /ʔabiɣiˈtiːdü/ abigitiidü
LA65.075 RB /poˈʔodü/ po’odü
LA65.075 RB /poʔoˈnübü/ po’onübü
LA65.076 RB /naɣaˈdiːka/ nagadiika
LA65.076 RB /tüˈvinda/ tüvinda
LA65.076 RB /saˈmeːna/ sameena
LA65.076 RB /tiːˈɣana/ tiigana
LA65.076 RB /pünikeːˈdiːka/ pünikeediika
LA65.076 RB /taʔinˈdiːna/ ta’indiina
LA65.076 RB /nahajeʔeˈkweːdü/ nahaye’ekweedü
LA65.076 RB /kaːˈdümü/ kaadümü
LA65.076 RB /pohoˈɣadü/ pohogadü
LA65.076 RB /pohoˈweːna/ pohoweena
LA65.076 RB /kaːpütiːkweːˈdiːna/ kaapütiikweediina
LA65.076 RB /mataˈsukwi/ matsukwi
LA65.076 RB /muɣuwaˈʔatü/ muguwa’atü
LA65.076 RB /taˈwabi/ tawabi
LA65.076 RB /pakaˈɣüdü/ pakagüdü
LA65.076 RB /naɣaviˈʔüdü/ nagavi’üdü 362
LA65.076 RB /pakaɣiˈʔidü/ pakagi’idü
LA65.076 RB /joʔoroˈkweːpü/ yo’orokweepü
LA65.076 RB /leːˈtseʔe/ leeche’e
LA65.076 RB /poroˈʔonü/ poro’onü
LA65.076 RB /püːwarüˈʔodü/ püüwarü’odü
LA65.076 RB /huveʔeˈʔüdü/ huve’e’üdü
LA65.076 RB /ˈtiːpü/ tiipü
LA65.077 RB /tsapuɣwiˈʔidü/ tsapugwi’idü
LA65.077 RB /ʔaˈwoho/ awoho
LA65.077 RB /ˈʔilu/ ilu
LA65.077 RB /ˈʔilu/ ilu
LA65.077 RB /mahaˈvadü/ mahavadü
LA65.077 RB /toˈtsaːtü/ totsaatü
LA65.077 RB /toˈtsaːtü/ totsaatü
LA65.077 RB /totsaːˈvistü/ totsaavishtü
LA65.077 RB /mahavaːˈdiːka/ mahavaadiika
LA65.077 RB /naˈvojo/ navoyo
LA65.077 RB /naˈvojo/ navoyo
LA65.077 RB /pütsakiˈneːka/ pütsakineeka
LA65.077 RB /koʔoˈvaːdü/ ko’ovaadü
LA65.077 RB /koʔoˈvaːdü/ ko’ovaadü 363
LA65.077 RB /totsivaʔaːküraˈvaːdü/ totsiva’aaküravaadü
LA65.077 RB /koʔovaːˈdiːka/ ko’ovaadiika
LA65.077 RB /küraˈvaːdü/ küravaadü
LA65.077 RB /küraˈnübü/ küranübü
LA65.077 RB /küraˈnübü/ küranübü
LA65.077 RB /tonˈdiːna/ tondiina
LA65.077 RB /taviˈnübü/ tavinübü
LA65.077 RB /tapitsiˈniːka/ tapichiniika
LA65.077 RB /toʔokwaˈtiʔi/ to’okwati’i
LA65.077 RB /puˈɣuzi/ puguzi
LA65.077 RB /ˈtübi/ tübi
LA65.077 RB /ˈkeːvi/ keevi
LA65.077 RB /weːˈnidü/ weenidü
LA65.077 RB /ˈtavi/ tavi
LA65.077 RB /tavikaːˈdiːka/ tavikaadiika
LA65.077 RB /ʔuːtsuˈkwasu/ uutsukwasu
LA65.077 RB /moʔosüˈkidü/ mo’osükidü
LA65.077 RB /ˈmüazi/ müazi
LA65.077 RB /kunaˈɣadü/ kunagadü
LA65.077 RB /huˈkubü/ hukubü
LA65.077 RB /huˈkubü/ hukubü 364
LA65.077 RB /ˈneːdü/ needü
LA65.077 RB /toːˈhovü/ toohovü
LA65.077 RB /mutsuˈneːdü/ mutsuneedü
LA65.077 RB /ʔaɣaɣwiˈsadü/ agagwisadü
LA65.077 RB /tazanoˈʔorü/ tazano’orü
LA65.077 RB /ʔuˈwarü/ uwaru
LA65.077 RB /ʔuˈwarü/ uwaru
LA65.077 RB /nuˈvavi/ nuvavi
LA65.078 RB /taʔasikweːˈvaːdü/ ta’asikweevaadü
LA65.078 RB /situʔipidüˈvaːdü/ situ’ipidüvaadü
LA65.078 RB /ˈtomo/ tomo
LA65.078 RB /tuˈɣunu/ tugunu
LA65.078 RB /tuɣunutürüˈɣavü/ tugunutürügavü
LA65.078 RB /ˈküawe/ küawe
LA65.078 RB /pükeːvaːˈdiːna/ pükeevaadiina
LA65.078 RB /ˈʔüːvi/ üüvi
LA65.078 RB /ʔuːtsuˈkwasu/ uutsukwasu
LA65.078 RB /ʔüːˈvisu/ üüvisu
LA65.078 RB /ˈtavi/ tavi
LA65.078 RB /ʔataːsiˈnija/ ataasiniya
LA65.078 RB /manüˈɣiju/ manügiyu 365
LA65.078 RB /waˈhatsa/ wahatsa
LA65.078 RB /navaˈhaju/ navahayu
LA65.078 RB /ʔaˈwatü/ awatü
LA65.078 RB /haboˈjoko/ haboyoko
LA65.078 RB /tüɣühapaˈdaːka/ tügühapadaaka
LA65.078 RB /paʔatoˈɣwidü/ pa’atogwidü
LA65.078 RB /ˈtoːvü/ toovü
LA65.078 RB /toˈɣowa/ togowa
LA65.078 RB /joziˈküdü/ yoziküdü
LA65.078 RB /ʔusaˈkweːdü/ usakweedü
LA65.078 RB /müziˈkweːdü/ müzikweedü
LA65.078 RB /pidüˈvaːdü/ pidüvaadü
LA65.078 RB /ʔusaˈkweːdü/ usakweedü
LA65.078 RB /pidüˈvaːdü/ pidüvaadü
LA65.078 RB /tünakweːvaːˈdiːna/ tünakweevaadiina
LA65.078 RB /jozikweːˈvaːdü/ yozikweevaadü
LA65.078 RB /tsipiːˈvaːdü/ tsipiivaadü
LA65.078 RB /tiːkweːˈvaːdü/ tiikweevaadü
LA65.079 RB /tsüɣütsüɣüvaˈdiːna/ chügüchüguvaadiina
LA65.079 RB /tiʔiˈdüwü/ ti’idüwü
LA65.079 RB /türüɣaːˈvika/ türügaavika 366
LA65.079 RB /türüɣaːˈvika/ türügaavika
LA65.079 RB /türüɣaːˈvika/ türügaavika
LA65.079 RB /ʔinaˈɣapü/ inagapü
LA65.079 RB /ˈsuːju/ suuyu
LA65.079 RB /puhiˈɣidü/ puhigid
LA65.079 RB /puhiˈɣidü/ puhigid
LA65.079 RB /puhiˈɣidü/ puhigid
LA65.079 RB /ʔamariˈjuʔu/ amariyu’u
LA65.079 RB /puhiˈɣidü/ puhigid
LA65.079 RB /ʔodoˈkwidü/ odokwidü
LA65.079 RB /tuhuˈkwidü/ tuhukwidü
LA65.079 RB /puhiˈɣidü/ puhigid
LA65.079 RB /tosoˈkwidü/ tosokwidü
LA65.079 RB /ʔotokwidütotsiˈɣadü/ otokwidütotsigadü
LA65.079 RB /puhiˈɣidü/ puhigid
LA65.079 RB /seːˈɣidü/ seegid
LA65.079 RB /tuhuˈkwidü/ tuhukwidü
LA65.079 RB /paʔatoˈɣodü/ pa’atogodü
LA65.079 RB /paʔatoɣoˈtiːna/ pa’atogotiina
LA65.079 RB /paʔatoˈɣodü/ pa’atogodü
LA65.079 RB /paˈʔadü/ pa’adü 367
LA65.079 RB /toveʔeˈpiːtsi/ tove’epich
LA65.079 RB /toveʔepiːtsiˈtiːna/ tove’epichtiina
LA65.079 RB /mutsuˈɣwidü/ mutsugwidü
LA65.079 RB /ʔaːsübiˈɣidü/ aasübigidü
LA65.079 RB /paʔatoˈɣodü/ pa’atogodü
LA65.079 RB /ʔohojeʔeˈkweːpü/ ohoye’ekweepu
LA65.079 RB /ʔohojeʔeˈkweːpü/ ohoye’ekweepü
LA65.079 RB /mohowaˈɣadü/ mohowagadü
LA65.079 RB /pijaˈviːna/ piyaviina
LA65.079 RB /ʔeːˈpizi/ eepizh
LA65.079 RB /ʔeːˈvizi eevizh
LA65.079 RB /ʔaˈsa:zi/ asaazi
LA65.079 RB /maʔaˈpüzi/ ma’apüzi
LA65.079 RB /ʔohoːˈvistü/ ohoovistü
LA65.079 RB /piˈjaːtü/ piyaatü
LA65.080 RB /ˈʔüːvi/ üüvi
LA65.080 RB /ˈʔüːvi/ üüvi
LA65.081 RB /ˈtiːpü/ tiipü
LA65.081 RB /pawahaˈɣadü/ pawahagadü
LA65.081 RB /keːˈvitsi/ keevichi
LA65.081 RB /ˈkeːvi/ keevi 368
LA65.081 RB /tüˈɣahni/ tügahni
LA65.081 RB /jüˈsüvü/ yüsüvü
LA65.081 RB /kuˈtsapü/ kutsap
LA65.081 RB /hoˈrodü/ horodü
LA65.081 RB /ʔüˈʔadü/ ü’adü
LA65.081 RB /jüˈhüdü/ yühüdü
LA65.081 RB /seːˈɣidü/ seegid
LA65.081 RB /wiʔaˈbitü/ wi’abitü
LA65.081 RB /ˈkuna/ kuna
LA65.081 RB /kwiˈhipi/ kwihipi
LA65.081 RB /kuˈtsapü/ kutsap
LA65.081 RB /hiɣwaˈtiːdü/ higwatiidü
LA65.081 RB /tuɣwiˈkweːdü/ tugwikweedü
LA65.081 RB /tuɣwitiːˈkweːna/ tugwitiikweena
LA65.081 RB /huɣwitiːnüˈvaːdü/ hugwitiinüvaadü
LA65.081 RB /noɣotseʔeˈkweːdü/ nogotse’ekweedü
LA65.081 RB /taruʔiˈdüsa/ taru’idüsa
LA65.081 RB /naruɣwaːʔiˈkweːpü/ narugwaa’ikweepü
LA65.081 RB /naruɣwaːʔiˈkweːpü/ narugwaa’ikweepü
LA65.081 RB /tavikaːˈriːdü/ tavikaariidü
LA65.081 RB /kiʔiˈvaːdü/ ki’ivaadü 369
LA65.081 RB /kiʔiˈdümü/ ki’idümü
LA65.081 RB /tukuːˈmüːtsi/ tukuumüüts
LA65.081 RB /tüˈhüja/ tühüya
LA65.081 RB /nohoponˈzinü/ nohoponzhinü
LA65.081 RB /kaˈʔadü/ ka’adü
LA65.081 RB /kaʔaˈvaːdü/ ka’avaadü
LA65.081 RB /kaʔaˈvaːdü/ ka’avaadü
LA65.081 RB /kaʔaˈdümü/ ka’adümü
LA65.081 RB /tümakaˈʔadü/ tümaka’adü
LA65.081 RB /tümakaˈʔadü/ tümaka’adü
LA65.081 RB /ʔataːˈkweːpü/ ataakweepü
LA65.081 RB /ʔataːˈkweːpü/ ataakweepü
LA65.081 RB /tüɣüjeˈʔedü/ tügüye’edü
LA65.081 RB /tüɣüjeˈʔedü/ tügüye’edü
LA65.081 RB /haːˈnizi/ haanizhi
LA65.081 RB /noˈpovü/ nopovü
LA65.081 RB /seːˈɣidü/ seegid
LA65.081 RB /keːˈsoʔo/ keeso’o
LA65.081 RB /hiˈvidü/ hividü
LA65.082 RB /hiviˈkweːna/ hivikweena
LA65.082 RB /sarˈveːsa/ sarveesa 370
LA65.082 RB /viˈnoʔo/ vino’o
LA65.082 RB /mansanahiˈveːka/ mansanahiveeka
LA65.082 RB /hoːˈkwidü/ hookwid
LA65.082 RB /hoːˈkwidü/ hookwid
LA65.082 RB /tüʔnijaˈnübü/ tü’niyanübü
LA65.082 RB /huˈwazi/ huwazi
LA65.082 RB /wiˈnapü/ winapü
LA65.082 RB /huˈwazi/ huwazi
LA65.082 RB /neheˈkapü/ nehekap
LA65.082 RB /kaˈrüdü/ karüdü
LA65.082 RB /kwidaˈtüa/ kwidatüa
LA65.082 RB /ˈkahni/ kahni
LA65.082 RB /ʔaɣaɣaːˈdiːka/ agagaadiika
LA65.082 RB /kahˈnija/ kahniya
LA65.082 RB /kaˈveːka/ kaveeka
LA65.082 RB /ventaˈnaʔa/ ventana’a
LA65.082 RB /nadawaˈnübü/ nadawanübü
LA65.082 RB /nadaˈwadü/ nadawadü
LA65.082 RB /laˈmeːsa/ lameesa
LA65.082 RB /tahanüˈweːka/ tahanüweeka
LA65.082 RB /karüˈnübü/ karünübü 371
LA65.082 RB /haviˈnübü/ havinübü
LA65.082 RB /haviˈnübü/ havinübü
LA65.082 RB /nüvükeːˈnübü/ nüvükeenübü
LA65.082 RB /paraˈʔasü/ para’asü
LA65.082 RB /saʔampuˈʔaka/ sa’mpu’aka
LA65.082 RB /kahˈnija/ kahniya
LA65.082 RB /kahˈnija/ kahniya
LA65.082 RB /koviˈtüa/ kovitüa
LA65.082 RB /nohopoːˈnizi/ nohopoonizhi
LA65.082 RB /nohopoindüˈvaːdü/ nohopoindüvaadü
LA65.082 RB /ˈkamü/ kamü
LA65.082 RB /sidoɣoʔovaːˈdiːka/ sidogo’ovaadiika
LA65.082 RB /sidoɣoʔoˈɣadü/ sidogo’ogadü
LA65.082 RB /ˈkaɣi/ kagi
LA65.082 RB /ˈkaɣi/ kagi
LA65.082 RB /mataˈsivü/ matasivü
LA65.082 RB /kwasuˈbüzi/ kwasubüzi
LA65.082 RB /naʔaɣaˈnübü/ na’aganübü
372
REFERENCES Avelino, H. (2010). Acoustic and electroglottographic analyses of nonpathological, nonmodal
phonation. Journal of Voice 24, pp. 270-280.
Bascom, B.W. (1965). Proto-Tepiman (Tepehuan-Piman). Doctoral dissertation, University of
Washington.
Beckman, M.E. (1994). When is a syllable not a syllable? Ohio State University Working Papers
in Linguistics 44, pp. 50-69.
Beckman, J.N. (2004). On the status of CODA COND in phonology. International Journal of
English Studies 4, 105-134.
Bell, A. (1978). Syllabic consonants. In Greenberg, Joseph H. (ed.), Universals of Human
Language Volume 2: Phonology . Stanford: Stanford University Press. pp. 153-201.
Berinstein, A.E. (1979). A Cross-Linguistic Study on the Perception and Production of Stress.
UCLA Working Papers in Phonetics 47. University of California, Los Angeles.
Berry, J., and Moyle, M. (2011). Covariation among vowel height effects on acoustic measures.
Journal of the Acoustical Society of America 130, pp. 365-371.
Bessell, N.J. (1992) Towards a phonetic and phonological typology of post-velar articulation.
Doctoral dissertation, University of British Columbia.
Bijankhan, M., and Nourbaksh, M. (2009). Voice onset time in Persian initial and intervocalic
stop production. Journal of the International Phonetic Association 39, pp. 335-364.
Blevins, J. (2003). The phonology of Yurok glottalized sonorants: Segmental fission under
syllabification. International Journal of American Linguistics 69, pp. 371-396. 373
Boersma, P., and Weenink, D. (2017). Praat: Doing phonetics by computer [Computer program].
Version 6.0.28, retrieved 23 March 2017 from http://www.praat.org.
Booth, C.G. (1979). Postpositions as verbs in Kawaiisu. International Journal of American
Linguistics 45, pp. 245-250 .
Borroff, M.L. (2007). A landmark underspecification account of the patterning of glottal stop.
Doctoral dissertation, Stony Brook University.
Breen, G., and Pensalfini, R. (1999). Arrernte: A language with no syllable onsets. Linguistic
Inquiry 30, 1-25.
Campbell, L., and Muntzel, M.C. (1989). The structural consequences of language death. In
Dorian, N.C. (ed.), Investigating Obsolescence: Studies in Language Contraction and
Death , pp. 181-196.
Casali, R.F. (1996). Resolving hiatus. Doctoral dissertation, University of California, Los
Angeles.
Charney, J.O. (1993). A Grammar of Comanche . University of Nebraska Press: Lincoln,
Nebraska.
Cho, T., Jun, S-A., and Ladefoged, P. (2002). Acoustic and aerodynamic correlates of Korean
stops and fricatives. Journal of Phonetics 30, pp. 193-228.
Cho, T., and Ladefoged, P. (1999). Variation and universals in VOT: Evidence from 18
languages. Journal of Phonetics 27, pp. 207-229.
Chomsky, N., and Halle, M. (1968). The Sound Pattern of English . New York, New York:
Harper & Row. 374
Clements, G. N., and Keyser, S. J. (1983). CV Phonology: A Generative Theory of the Syllable.
Cambrige, Massachusetts: MIT Press.
Crapo, R.H. (1970). The origins of directional adverbs in Uto-Aztecan languages. International
Journal of American Linguistics 36, pp. 181-189.
Crum, B., and Dayley, J. (1993). Western Shoshoni Grammar . Boise, Idaho: Boise State
University Department of Anthropology.
Crystal, T.H., & House, A.S. (1988a). The duration of American-English vowels: An overview.
Journal of Phonetics 16, pp. 263-284.
Crystal, T.H., and House, A.S. (1988b). Segmental durations in connected-speech signals:
Current results. Journal of the Acoustical Society of America 83.4, pp. 1553-1573.
Dayley, J.P. (1989). Tümpisa (Panamint) Shoshone Grammar . Berkeley, California: University
of California Press. de Jong, K.J. (1991). The oral articulation of English stress accent. Doctoral dissertation, The
Ohio State University. de Jong, K.J. (1995). The supraglottal articulation of prominence in English: Linguistic stress as
localized hyperarticulation. Journal of the Acoustical Society of America 97, pp. 491-504.
Delattre, P. (1966). A comparison of syllable length conditioning among languages.
International Review of Applied Linguistics in Language Teaching 4.3 , pp. 183-199.
Delforge, A.M. (2008a). Unstressed vowel reduction in Andean Spanish. In Laura Colantoni and
Jeffrey Steele (eds), Selected Proceedings of the Third Conference on Laboratory
Approaches to Spanish Phonology , pp. 107-124. Somerville, Massachusetts: Cascadilla
Proceedings Project. 375
Delforge, A.M. (2008b). Gestural alignment constraints and unstressed vowel devoicing in
Andean Spanish. In Charles B. Chang and Hannah J. Haynie (eds), Proceedings of the
26 th West Coast Conference on Formal Linguistics, pp. 147-155. Somerville,
Massachusetts: Cascadilla Proceedings Project.
Dorian, N.C. (ed.) (1989). Investigating Obsolescence: Studies in Language Contraction and
Death. Cambridge: Cambridge University Press.
Doty, C.S., Idemaru, K., and Guion, S.G. (2007). Singleton and geminate stops in Finnish –
Acoustic correlates. In Interspeech 2007, 8 th Annual Conference of the International
Speech Communication Association, Antwerp, Belgium (pp.2737-2740).
Elzinga, D.A. (1999). The consonants of Gosiute. Doctoral dissertation, University of Arizona.
Fischer-Jørgensen, E. (1954). Acoustic analysis of stop consonants. Miscellanea Phonetica 2, pp.
42-59.
Fishman, J. (1991). Reversing Language Shift: Theoretical and Empirical Foundations of
Assistance to Threatened Languages . Multilingual Matters Ltd.: Philadelphia,
Pennsylvania.
Fourakis, M. (1991). Tempo, stress, and vowel reduction in American English. Journal of the
Acoustical Society of America 90, pp. 1816-1827.
Fowler, C.S. (1983). Some lexical cues to Uto-Aztecan prehistory. International Journal of
American Linguistics 49, pp. 224-257.
Freeze, R., and Iannucci, D.E. (1979). Internal classification of the Numic languages of Uto-
Aztecan. Amerindia 4, pp. 77-94. 376
Fry, D.B. (1955). Duration and intensity as physical correlates of linguistic stress. Journal of the
Acoustical Society of America 27.4, pp. 765-786.
Fry, D.B. (1958). Experiments in the perception of stress. Language and Speech 1, pp. 126-152.
Fudge, E.C. (1969). Syllables. Journal of Linguistics 5, 253-286.
Gay, T. (1978). Effect of speaking rate on vowel formant movements. Journal of the Acoustical
Society of America 63, pp. 223-230.
Gerfen, C. (1999). Phonology and phonetics in Coatzsopan Mixtec. Dordrecht: Kluwer
Academic Publishers
Gerfen, C., and Baker, K. (2005). The production and perception of laryngealized vowels in
Coatzospan Mixtec. Journal of Phonetics 33, pp. 311-334.
Gobl, C. (1989). A preliminary study of acoustic voice quality correlates. STL-QPSR 30.4, pp. 9-
22.
Goddard, I. (1983). Introduction. In Goddard, I. (ed)., Hamdbook of North American Indians .
Smithsonian Institute: Washington, D.C.: pp. 1-16.
Gordon, M., and Ladefoged, P. (2001). Phonation types: A cross-linguistic overview. Journal of
Phonetics 29, pp. 383-406.
Grant, L. (2018). Personal communication. 27 Mar 2018.
Green, L. (2013). The Kawaiisu language revitalization effort in southern California. SOAS
Working Papers in Linguistics 16, pp. 279-289.
Hammond, J. (2018, August 22). Pen in hand: Go Native Day: A rare chance to experience
native California culture. Tehachapi News , pp. 12, 19. Retrieved from
http://www.tehachapinews.com. 377
Hasegawa, Y. (1999). Pitch accent and vowel devoicing in Japanese. In J.J. Ohala, Y. Hasegawa,
M. Ohala, D. Granville, and A.C. Bailey (eds.), Proceedings of the XIVth International
Congress of Phonetic Sciences, San Francisco, 1-7 August 1999 , pp. 523-526.
Haugen, E. (1956). Syllabification in Kutenai. International Journal of American Linguistics 22,
196-201.
Haugen, J.D. (2008). Morphology at the Interfaces: Reduplication and Noun Incorporation in
Uto-Aztecan . John Benjamins North America: Philadelphia, Pennsylvania.
Hayes, Bruce. (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry 20,
253-306.
Hayward, K. (2000). Experimental Phonetics. Harlow: Longman.
Heath, J. (1977). Uto-Aztecan morphophonemics. International Journal of American Linguistics
43, pp. 27-36.
Heath, J. (1978). Uto-Aztecan *na- class verbs. International Journal of American Linguistics
44, pp. 211-222.
Heath, J. (1985). Proto-Northern Uto-Aztecan participles. International Journal of American
Linguistics 51, pp. 441-443.
Herrick, D. (2011). On Comanche’s central mid vowel. International Journal of American
Linguistics 77, pp. 373-396.
Hill, J.H. (2001). Proto-Uto-Aztecan: A community of cultivators in central Mexico? American
Anthropologist 103, pp. 913-934.
Hill, J.H. (2008). Northern Uto-Aztecan and Kiowa-Tanoan: Evidence of contact between the
proto-languages? International Journal of American Linguistics 74, pp.155-158. 378
Hill, J.H., and Hill, K.C. (1970). A note on Uto-Aztecan color terminologies. Anthropological
Linguistics 12.7, pp. 231-238.
Hinton, L. (1994). Flutes of Fire: Essays on California Indian Languages. Heyday Books:
Berkeley, California.
Hinton, L. (2002). How to Keep Your Language Alive: A Commonsense Approach to One-on-
One Language Learning. Heyday Books: Berkeley, California.
Hockett, C. (1955). A Manual of Phonology. Baltimore: Waverly Press.
Holman, E.W., Brown, C.H., Wichmann, S., Müller, A., Velupillai, V., Hammarström, H., Jung,
H., Bakker, D., Brown, P., Belyaev, O., Urban, M., Mailhammer, R., List, J., and Egorov,
D. (2011). Automated dating of the world’s language families based on lexical similarity.
Current Anthropology , pp. 841-875.
Hooper, J.B. (1978). Constraints on schwa deletion in American English. In Jacek Fisiak (ed),
Recent Developments in Historical Phonology , pp. 183-208. The Hague: Mouton
Publishers.
Hughes, A.J. (2016). Akimel O’otham Reading and Writing: A New Beginning. M.A. Thesis,
University of Arizona.
Iannucci, D. E. (1973). Numic historical phonology. Doctoral dissertation, Cornell University.
Ito, J. (1987). Syllable theory in prosodic phonology. Doctoral dissertation, University of
Massachusetts.
Ito, J. (1989). A prosodic theory of epenthesis. Natural Language & Linguistic Theory 7, 217-
259. 379
Jakobson, R., Fant, C.G.M., and Halle, M. (1952). Preliminaries to Speech Analysis: The
Distinctive Features and their Correlates. Cambridge, Massachusetts: MIT Press.
Johnson, K. (2004). Massive reduction in conversational American English. In K. Yoneyama and
K. Maekawa (eds), Spontaneous Speech: Data and Analysis Proceedings of the 1 st
Session of the 10 th International Symposium , pp. 29-54. Tokyo, Japan: The National
International Institute for Japanese Language.
Johnson, K. (2012). Acoustic and Auditory Phonetics. Malden, Massachusetts: Wiley-Blackwell.
Johnson, K.E. (2008). Second language acquisition of the Spanish multiple vibrant consonant.
Doctoral dissertation, University of Arizona.
Jongman, A., Wayland, R., and Wong, S. (2000). Acoustic characteristics of English fricatives.
Journal of the Acoustical Society of America 108, p. 1252.
Kawagoe, I. (2002). Acquisition of English word stress by Japanese learners. In J.M. Liceras, H.
Sobi, and H. Goodluck (eds.), L2 Links: Proceedings of the 2002 Generative Approaches
to Second Language Acquisition Conference. Ottawa, Canada.
Kawaiisu Language & Cultural Center (2018). [Website of the Kawaiisu Language & Cultural
Center]. Retrieved from http://www.kawaiisu.org.
Keating, P., Cho, T., Fougeron, C., and Hsu, C-S. (2003). Domain-initial articulatory
strengthening in four languages. In J. Local, R. Ogden, and R. Temple (eds.), Phonetic
Interpretation: Papers in Laboratory Phonology VI, pp. 145-163 .
Kent, R.D., and Netsell, R. (1971). Effects of stress contrasts on certain articulatory parameters.
Phonetica 24, pp. 23-44. 380
Kidder, E. (2008). Tone, intonation, stress and duration in Navajo. Coyote Papers: Working
Papers in Linguistics, Linguistic Theory at the University of Arizona 16, pp. 55-66.
Kidder, E. (2013). Prominence in Yucatec Maya: The role of stress in Yucatec Maya words.
Doctoral dissertation, University of Arizona.
Klein, S. (1958, 1981-1983). The Sheldon Klein collection of Kawaiisu sound recordings. LA
65. Berkeley Language Center, University of California, Berkeley.
< http://cla.berkeley.edu/collection/10140 >
Klein, S. (1959). Comparative Mono-Kawaiisu. International Journal of American Linguistics
25, pp. 233-238.
Klein, S. (2002). Tying loose ends in Kawaiisu phonology. Some comments on Zigmond, Booth,
and Munro (1990). In Report #12: Proceedings of the 50 th Anniversary Conference of the
Survey of California and Other Indian Languages, Lisa Conathan and Teresa McFarland,
eds. Berkeley, California.
Koopmans-van Beinum, F.J. (1980). Vowel contrast reduction: An acoustic and perceptual study
of Dutch vowels in various speech conditions. Doctoral dissertation, University of
Amsterdam.
Ladefoged, P. (1963). Some physiological parameters in speech. Language & Speech 6, pp. 109-
119.
Ladefoged, P. (2003). Phonetic Data Analysis: An Introduction to Fieldwork and Instrumental
Techniques. Blackwell Publishing: Malden, Massachusetts.
Ladefoged, P. (2006). A Course in Phonetics. Boston, Massachusetts: Thomson Wadsworth.
Lamb, S.M. (1957). Mono grammar. Doctoral dissertation, University of California. 381
Lamb, S.M. (1958). Linguistic prehistory in the Great Basin. International Journal of American
Linguistics 24, pp. 95-100.
Langacker, R.W. (1970). The vowels of Proto Uto-Aztecan. International Journal of American
Linguistics 36, pp. 169-180.
Langacker, R.W. (1976a). A note on Uto-Aztecan consonant gradation. International Journal of
American Linguistics 42, pp. 374-379.
Langacker, R.W. (1976b). Non-distinct arguments in Uto-Aztecan. University of California
Publications in Linguistics 82. University of California Press: Berkeley, California.
Lawyer, L.C. (2015). Patwin phonemics, phonetics, and phonotactics. International Journal of
American Linguistics 81, pp. 221-260.
Lewis, M.P., and Simons, G.F. (2010). Assessing endangerment: Expanding Fishman’s GIDS.
Revue Roumaine de Linguistique 55, pp. 103-120.
Lieberman, P. (1960). Some acoustic correlates of word stress in English. Journal of the
Acoustical Society of America 32.4, pp. 451-454.
Lieberman, P. (1967). Intonation, Perception, and Language. Cambridge, Massachusetts: MIT
Press.
Lindblom, B. (1963). Spectrographic study of vowel reduction. Journal of the Acoustical Society
of America 35, pp. 1773-1781.
Lindblom, B. (1967). Vowel duration and a model of lip mandible coordination. Speech
Transmission Laboratory Quarterly Progress Report 4, pp. 1-29.
Lisker, L., and Abramson, A. (1964). A cross-language study of voicing in initial stops:
Acoustical measurements. Word 20, pp. 384-422. 382
Lisker, L., and Abramson, A. (1967). Some effects of context on voice onset time in English
stops. Language and Speech 10, pp. 1-28.
Macaulay, M., and Salmons, J. (1995). The phonology of glottalization in Mixtec. International
Journal of American Linguistics 61, pp. 38-61.
Manaster-Ramer, A. (1984). Kern laws. International Journal of American Linguistics 50, pp.
325-334.
Manaster-Ramer, A. (1993). On lenition in some Northern Uto-Aztecan languages. International
Journal of American Linguistics 59, pp. 334-341.
McCawley, J.D. (1968). The Phonological Component of a Grammar of Japanese. Mouton: The
Hague.
McDonough, J. (2003). The Navajo Sound System. Dordrecht: Kluwer Academic Publishing.
Miller, W.R. (1966). Anthropological linguistics in the Great Basin. In d’Azevedo, W.L. et al.
(eds)., The Current Status of Anthropological Research in the Great Basin , pp. 75-112.
Desert Research Institute: Reno, Nevada.
Miller, W.R. (1967). Uto-Aztecan cognate sets. University of California Publications in
Linguistics, Vol. 48. University of California Press: Berkeley, California.
Miller, W.R. (1983). Uto-Aztecan languages. In Goddard, I. (ed)., Hamdbook of North American
Indians . Smithsonian Institute: Washington, D.C.: pp. 113-124.
Miller, W.R. (1984). The classification of the Uto-Aztecan languages based on lexical evidence.
International Journal of American Linguistics 50, pp. 1-24.
Miller, W.R., Elzinga, D., and McLaughlin, J.E. (2005). Preaspiration and gemination in Central
Numic. International Journal of American Linguistics 71, pp. 413-444. 383
Mithun, M. (1999) The Languages of Native North America. University of Cambridge Press:
Cambridge.
Morton, J., and Jassem, W. (1965). Acoustic correlates of stress. Language and Speech 8.3, pp.
159-181.
Munro, P. (1977). Towards a reconstruction of Uto-Aztecan stress. In Hyman, L.M. (ed). Studies
in Stress and Accent , pp. 303-326. Southern California Occasional Papers in Linguistics
4. Department of Linguistics, University of Southern California: Los Angeles, California.
Nguyen, T.T.A,. and Ingram, J. (2005). Vietnamese acquisition of English word stress. TESOL
Quarterly 39.2, pp. 309-319.
Nichols, M. (1974). Northern Paiute historical grammar. Doctoral dissertation: University of
California, Berkeley.
O’Connor, J.D., and Trim, J. L. M. (1953). Vowel, consonant, and syllable – A phonological
definition. Word 9, 103-122.
Oberly, S. I. (2008). A phonetic analysis of Southern Ute with a discussion of Southern Ute
language policies and revitalization. Doctoral dissertation, University of Arizona.
Oberly, S.I., and Kharlamov, V. (2015). The phonetic realization of devoiced vowels in the
Southern Ute language. Phonetica 72, pp. 1-19.
Ogasawara, N. (2007). Processing of speech variability: Vowel reduction in Japanese. Doctoral
dissertation, University of Arizona.
Ohala, J.J. (1977). The physiology of stress. In L.M. Hyman (ed.), Studies in Stress and Accent:
Southern California Occasional Papers in Linguistics No. 4 , p. 145. 384
Ohala, J.J. (1983). The origin of sound patterns in vocal tract constraints. In P.F. MacNeilage
(ed.), The Production of Speech, pp. 189-216 . New York, Springer-Verlag.
Ohala, J.J. (1997). Aerodynamics of phonology. Proceedings of the 4 th Seoul International
Conference on Linguistics 11-15 Aug 1997 , pp. 92-97.
Ohala, M., and Ohala, J.J. (1992). Phonetic universals and Hindi segment duration. In Ohala, J.J.,
Nearey, T., Derwing, B., Hodge, M., and Wiebe, G. (eds.), Proceedings of the
International Conference on Spoken Language Processing. Edmonton: University of
Alberta, pp. 831-834.
Ohala, J.J., and Riordan, C.J. (1979). Passive vocal tract enlargement during voiced stops. In J.J.
Wolf and D.H. Klatt (eds.), Speech Communication Papers, pp. 89-92. New York:
Acoustical Society of America.
Ohala, J.J., Riordan, C.J., and Kawasaki, H. (1979). Investigation of pulmonic activity in speech.
In E.J. Fischer-Jørgensen, J. Rischel, and N. Thorsen (eds.), Proceedings of the Ninth
International Congress of Phonetic Sciences, Volume I , p. 205 . Institute of Phonetics,
University of Copenhagen.
Oller, D.K. (1973). The effect of position in utterance on speech segment duration in English.
Journal of the Acoustical Society of America 54, pp.1235-1247 .
Oller, D.K., and Smith, B.L. (1977). Effect of final-syllable position on vowel duration in infant
babbling. Journal of the Acoustical Society of America 62, pp. 994-997.
Peterson, G.E., and Lehiste I. (1960). Duration of syllable nuclei in English. Journal of the
Acoustical Society of America 32, pp. 693-703. 385
Plug, L., and Ogden, R. (2003). A parametric approach to the phonetics of postvocalic /r/ in
Dutch. Phonetica 60, pp. 159-186.
Press, M.L. (1979). Chemehuevi: A grammar and lexicon. Berkeley, California: University of
California Press.
Pullum, G.K., and Ladusaw, W.A. (1996). Phonetic Symbol Guide. Chicago, Illinois: The
University of Chicago Press.
Quilis, A. (1981). Fonética acústica de la lengua Española. Madrid: Gredos.
Romney, A.K. (1957). The genetic model and Uto-Aztecan time perspective. Davidson Journal
of Anthropology 3, pp. 35-41.
Sagey, E. (1986). The representation of features and relations in non-linear phonology. Doctoral
dissertation, MIT.
Sapir, E. (1931). The Southern Paiute Language, Parts I, II, and III. AAASP 65. 1-730.
Sapir, E. (1949). The psychological reality of phonemes. In Mandelbaum, David (ed.), Selected
Writings of Edward Sapir , 46-60. University of California Press: Berkeley.
Saussure, Ferdinand de. (1922). A Course in General Linguistics. New York: Philosophical
Library.
Saxton, D., Saxton, L., and Enos, S. (1983). Tohono O’odham/Pima to English English to
Tohono O’odham Pima Dictionary. Tucson, Arizona: University of Arizona Press.
Schuppler, B., Ernestus, M., Scharenbourg, O., and Boves, L. (2011). Acoustic reduction in
conversational Dutch: A quantitative analysis based on automatically generated
segmental transcriptions. Journal of Phonetics 39, pp. 96-109. 386
Shaul, D.L. (1983). The position of Opata and Eudeve in Uto-Aztecan. Studies in Native
American Languages II: Kansas Working Papers in Linguistics 8.2 , pp. 95-123.
Sherman, D. (1975). Stop and fricative systems: A discussion of paradigmatic gaps and the
question of language sampling. Stanford Working Papers in Language Universals 17, pp.
1-31.
Silva, D.J. (1997). The variable deletion of unstressed vowels in Faialense Portuguese. Language
Variation and Change 9.3, pp. 295-308.
Silva, W. (2016). The status of the laryngeals ‘ʔ’ and ‘h’ in Desano. In H. Avelino, M. Coler, and
W.L. Wetzels (eds.), The Phonetics and Phonology of Laryngeal Features in Native
American Languages , pp. 285-307. Leiden: Brill.
Silver, S., and Miller, W.R. (1997). American Indian Languages: Cultural and Social Contexts .
University of Arizona Press: Tucson, Arizona.
Simons, G.F., and Fennig, C.D. (eds.) (2017). Ethnologue: Languages of the World, Twentieth
Edition. SIL International: Dallas, Texas.
Steele, S. (1973). Futurity, intention, and possibility: A semantic reconstruction in Uto-Aztecan.
Papers in Linguistics 6, pp. 1-37.
Steele, S. (1975). Past and irrealis: Just what does it all mean? International Journal of American
Linguistics 41, pp. 200-217.
Stubbs, B.D. (2000). More palatable reconstructions for Uto-Aztecan palatals. International
Journal of American Linguistics 66, pp. 125-137.
Stubbs, B.D. (2011). Uto-Aztecan: A Comparative Vocabulary . Rocky Mountain Books and
Productions: Blanding, Utah. 387
Taylor, W.W. (1961). Archaeology and language in western North America. American Antiquity
27, pp. 71-81.
Thomas, P.N. (2017). The central vowel of Kawaiisu. International Journal of American
Linguistics 83, pp. 539-559.
Tiffany, W.R. (1959). Nonrandom sources of variation in vowel quality. Journal of Speech,
Language, and Hearing Research 2, pp. 305-317.
Toosarvandani, M. (2010). Patterns of nominalization in Numic. International Journal of
American Linguistics 76, pp. 71-100.
Torreira, F., and Ernestus, M. (2011). Vowel elision in casual French: The case of vowel /e/ in
the word c’était. Journal of Phonetics 39, pp. 50-58.
Tucker, B.V. (2007). Acoustic phonetic description of Chemehuevi. Presented at the annual
meeting of the Society for the Study of Indigenous Languages of America (Anaheim,
California).
Tucker, B.V. (2013). An acoustic description of Chemehuevi. Presented at the 166 th Meeting of
the Acoustical Society of America (San Francisco, CA).
US Census Bureau (2010). American Indian and Alaska Native Tribes in the United States and
Puerto Rico: 2010 . Retrieved from
https://www.census.gov/population/www/cen2010/cph-t/t-6tables/TABLE%20(1).pdf.
Van Bezooijen, R., Kroezen, S., and van den Berg, R. (2002). Front approximant /r/: A new and
vigorous change in Dutch. Linguistics in the Netherlands 19, pp. 1-11. 388
Voegelin, C.F., Voegelin, F.M., and Hale, K.L. (1962). Typological and comparative grammar of
Uto-Aztecan. Indiana University Publications in Anthropology and Linguistics. Memoir
17 of the International Journal of American Linguistics.
Wang, Q. (2008). L2 stress perception: The reliance on different acoustic cues. In P.A. Barbosa,
S. Madureira, and C. Reis (eds.), Proceedings of the Speech Prosody 2008 Conference.
May 6-9, 2008. Campinas, Brazil.
Warner, N., and Tucker, B.V. (2017). An effect of flaps on the fourth formant in English.
Journal of the International Phonetic Association 47 , pp. 1-15.
Whitney, W. (1874). Oriental and Linguistic Studies – Second Series. New York: Scribner.
Yamada, R. (2011). Integrating documentation and formal teaching of Kari’nja: Documentary
materials as pedagogical materials. Language Documentation and Conservation 5, pp. 1-
30.
Zepeda, O. (1983). A Tohono O’odham Grammar. Tucson, Arizona: University of Arizona
Press.
Zigmond, M.L., Booth, C.G., and Munro, P. (1990). Kawaiisu: A grammar and dictionary with
texts. Berkeley, California: University of California Press.