THE PHONOLOGY OF PROTO-TAI A Dissertation Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy by Pittayawat Pittayaporn August 2009 © 2009 Pittayawat Pittayaporn THE PHONOLOGY OF PROTO-TAI Pittayawat Pittayaporn, Ph. D. Cornell University 2009 Proto-Tai is the ancestor of the Tai languages of Mainland Southeast Asia. Modern Tai languages share many structural similarities and phonological innovations, but reconstructing the phonology requires a thorough understanding of the convergent trends of the Southeast Asian linguistic area, as well as a theoretical foundation in order to distinguish inherited traits from universal tendencies, chance, diffusion, or parallel development. This dissertation presents a new reconstruction of Proto-Tai phonology, based on a systematic application of the Comparative Method and an appreciation of the force of contact. It also incorporates a large amount of dialect data that have become available only recently. In contrast to the generally accepted assumption that Proto-Tai was monosyllabic, this thesis claims that Proto-Tai was a sesquisyllabic language that allowed both sesquisyllabic and monosyllabic prosodic words. In the proposed reconstruction, it is argued that Proto-Tai had three contrastive phonation types and six places of articulation. It had plain voiceless, implosive, and voiced stops, but lacked the aspirated stop series (central to previous reconstructions). As for place of articulation, Proto-Tai had a distinctive uvular series, in addition to the labial, alveolar, palatal, velar, and glottal series typically reconstructed. In the onset, these consonants can combine to form tautosyllabic clusters or sequisyllabic structures. Regarding the rime, PT had seven vowel qualities that contrasted in height, backness, and rounding. A vowel length contrast also existed for each quality. Palatal and lateral consonants also occurred in the coda in addition to the final consonants generally assumed. Furthermore, Proto-Tai was a tonal language whose four tonal categories *ABCD contrasted both in terms of pitch and voice quality. Many of these Proto-Tai traits are not attested in modern Tai languages. The current reconstruction of Proto-Tai phonology is thus a demonstration of the power of the Comparative Method as well as the role that phonological theory can play in reconstruction. This thesis presents a picture of the history of Tai languages as characterized by divergent changes overridden by waves of convergent trends that transformed Proto-Tai into a network of typologically homogenous dialects that differ markedly from their parent. This analysis offers a comprehensive account of the transformation of the Proto-Tai phonology into modern systems. BIOGRAPHICAL SKETCH Pittayawat Pittayaporn, known among friends as Joe, attended Sarawittaya High School, Bangkok, before entering the Faculty of Arts at Chulalongkorn University, in 1998. He graduated with a Bachelor of Arts in Spanish (1st class honor) in 2002. iii pācerācariyā honti guṇuttarānusāsakā paññāvuḍḍhikare te te dinnovāde namāmihaṃ Teachers and teachers upon teachers are superior instructors because of their merits. I bow to those who provide instruction; who inspire knowledge and prosperity. iv ACKNOWLEDGMENTS Writing this dissertation was a long process full of mysteries and discoveries, as well as smiles, distresses, laughter. This dissertation represents moments shared with so many people whom this research brought into my life, and for whom I am deeply grateful. Above all, I am greatly indebted to my beloved family—my father Sanan, my mother Issayaporn, my brother Norrasate, and my aunt Manee—for their love and sacrifices made me who I am. I would like to thank them for their support and incredible patience throughout my life. I offer this dissertation as a token of my utmost respect and deepest gratitude toward all my teachers. It is my greatest honor and privilege to be a student of my thesis advisor Professor John Whitman. He is a careful, thorough, and inspiring scholar, as well as a caring, attentive, and generous mentor. Words cannot describe how much I value those late evening meetings, those lunch discussions, those letters of recommendations, and above all the friendship he has offered. I am greatly indebted to Professor Pranee Kullavanijaya, who introduced me Tai historical linguistics, and has been a teacher to me in the truest sense of the word. I would like to express my appreciation for Professor Theraphan Luangthongkum, who imparted to me her incredible knowledge on Southeast Asian Linguistics, and who inspired me to be a good field linguist. I am tremendously indebted to Professor MR. Kalaya Tingsabadh, who has always been a respected teacher and benefactor to me. Without her kindness, my life would be very different. This dissertation is also an homage to Fang-kuei Li, William Gedney, André-Georges Haudricourt, Paul Benedict, and other scholars before me whose invaluable contributions made this dissertation possible. I am indebted to all my language consultants who generously and patiently worked with me: Pan and Jai Kaewkhamsaen of Waritchapum; Dao Van Vinh, and v Hoang Thi Du of Sapa; Hoang Thi Cu of Bao Yen and family; Prof. Hoang Van Ma; Li Jingsen, You Zhouhui of Shangsi and their families; and Chen Yufang and Chen Yujiao of Jinxiu. My experience in the field was a memorable one. I was especially touched by the generosity and kindness of Ajarn Pan Kaewkhamsaen, Hoang Thi Cu, Zhouhui, and their families for their warm welcome to their home. In addition, I thank the Department of Linguistics, Vietnam National University, Hanoi; the Office of Culture and Information, Lao Cai Province; and the School of Foreign Studies, Guangxi University for Nationalities, without whose assistance my fieldwork would not have been possible. I am also very grateful to my friend Marc Brunelle who was there at the beginning and helped bring my field research in Vietnam to light. I would like to express my appreciation to my thesis committee members Prof. Abigail Cohn, Prof. Michael Weiss, and Prof. Draga Zec for their insightful comments, valuable suggestions, and helpful advice. Moreover, this dissertation would not be possible without generous help from so many people at various stages of my Ph.D. experience: Prof. John Wolff, Prof. James Matisoff, Prof. Michel Ferlus, Prof. Laurent Sagart, Prof. Maneepin Phromsuthirak, Prof. Prapod Assavavirulhakarn, Prof. Thak Chaloemtiarana, Prof. Kingkarn Thepkanjana, Prof. Somsongse Burusphat, Prof. Mitsuaki Endo, Prof. Jerrold Edmondson, Prof. Anthony Diller, Prof. Tran Tri Doi, Prof. Zhang Youjun, Prof. Fang Yin, Prof. Bo Wenze, Prof. Huang Xiulian, Dr. Tran Huu Son, Dr. Phichet Saiphan, Dr. Alexis Michaud, Dr. Peter Norquest, Dr. Weera Ostapirat, Komsorn Aukyapisut (P’Pop), Mongkol Tonphuban, Qin Xiuhong, You Huicai and family, Lu Sheng, Do Quang Son, Phan Luong Hung, Ryuichi Kosaka, Eric Johnson, Andrew Joseph, Sheila Haddad, Angie Tinti, Nancy Loncto, Wendy Treat, Gina Giambattista Cesari, Renee Milligan, Suwanna Charoensiri, Natcha Preechawattanasakul, Supaporn Phoprom, Nares Petcharat, Vu Huyen, Supawat Chomchan, Liu Futing, Zheng Wuyao, and numerous others. vi I also extend special thanks to all my friends for their help, moral support, and, above all, their friendship—Nawaporn Inkaew, Arthid Sheravanichkul, Youththachai Vitheekol, and my friends from the Faculty of Arts for their long-lasting friendship and for being my strength in these difficult years; Woraporn Sukhumawasi, Pimnara Hirankasi, Benjarong Suwankiri, Ariya Chiralertpong, Worawat Sangthongsakaw, and Neeranat Photiyanon, for their wonderful friendship and for always being there for me; Tejaswini Deoskar, John Duong Phan, Marc Brunelle, Ivan Small, Alice Iida, Pamela Corey, Eileen Vo, Johanna Brugman, Andrew Joseph, Dan Kaufman, Andrew Johnson, Daena Funahashi, Claudine Ang, Samson Lim, Nikola Predolac, and HyeSook Lee for their ever-growing friendship, and for making these Ithaca years of my life marvelous ones. One thing in life that I am most grateful for is that I have friends, who are wonderful and true. It is not possible to mention all of them on this piece of paper, but they will always be in my heart. I am indebted to Cornell’s Graduate School, Cornell’s East Asia Program, and the Lam Family Fund for South China Research for their generous financial support which made the this dissertation, especially the field research possible. I express my appreciation to the Department of Linguistics, Cornell University for all its support. It is this department that has nurtured me intellectually in these past six years. I am indebted to Cornell’s Southeast Asia Program not only for their financial support but also for being my “home” at Cornell. I express my deepest gratitude to the Anandamahidol Foundation under the Royal Patronage of His Majesty the King of Thailand for giving me the opportunity to pursue this Ph.D. study. I would like to pay my utmost respect and gratitude to His Majesty King Bhumibol Adulyadej and Her Royal Highness Princess Sirindhorn for their vision in recognizing the importance of education for the well-being of the people of Thailand. vii TABLE OF CONTENTS Bibliographical sketch……………………………………………………….………iii Dedication…………………………………………………………………………….iv Acknowledgments…………………………………………………………………….v List of figures……………………………………………………………………..…xiii List of tables…………………………………………………………………………xiv List of abbreviations………………………………………………………......……xix Chapter 1....................................................................................................................... 1  1.1  Introduction ......................................................................................................... 1  1.2  Tai languages....................................................................................................... 2  1.3  Conventional reconstruction of PT and pending issues ...................................... 7  1.3.1  Li’s reconstruction ........................................................................................... 7  1.3.2  Outstanding problems .................................................................................... 13  1.4  Methodology, data, and notation ....................................................................... 16  1.4.1  Methodology.................................................................................................. 16  1.4.2  Data and fieldwork......................................................................................... 18  1.4.2.1  Published materials .................................................................................... 20  1.4.2.2  Fieldwork ................................................................................................... 20  1.4.2.3  Representative dialects............................................................................... 21  1.4.2.4  Tai-external data......................................................................................... 21  1.4.3  Notation ......................................................................................................... 22  1.5  Overview of dissertation.................................................................................... 26  Chapter 2..................................................................................................................... 28  2.1  Introduction ....................................................................................................... 28  2.2  The prosodic word............................................................................................. 30  2.2.1  The importance of the prosodic word ........................................................... 31  2.2.2  Well-formedness ............................................................................................ 32  2.2.2.1  Size requirements ....................................................................................... 32  2.2.2.2  Sonority ...................................................................................................... 34  2.2.2.3  Foot typology ............................................................................................. 37  2.3  Reconstructing the PT prosodic word ............................................................... 39  2.3.1  The PT prosodic word in earlier proposals .................................................... 40  2.3.2  Sesquisyllabicity in PT .................................................................................. 43  2.4  The prosodic word in Southeast Asia................................................................ 52  2.4.1  Monosyllabic languages ................................................................................ 53  2.4.2  Disyllabic languages ...................................................................................... 56  2.4.3  Sesquisyllabic languages ............................................................................... 59  2.5  PT as a sesquisyllabic language ........................................................................ 63  2.5.1.1  Onset .......................................................................................................... 65  2.5.1.2  Rime ........................................................................................................... 66  2.5.1.3  Tone............................................................................................................ 67  2.6  Summary............................................................................................................ 68  viii Chapter 3..................................................................................................................... 69  3.1  Introduction ....................................................................................................... 69  3.2  Inventory of PT simple onsets........................................................................... 69  3.3  A richer dorsal series ......................................................................................... 72  3.4  “Voicing alternation” revisited.......................................................................... 77  3.5  Lack of aspiration in PT .................................................................................... 88  3.5.1  Li’s simple aspirates as post-PT innovations................................................. 89  3.5.2  Aspiration of original PT clusters .................................................................. 93  3.6  Reconstructions of PT simple initials.............................................................. 101  3.6.1  Voiceless stops............................................................................................. 101  3.6.1.1  *p- ............................................................................................................ 101  3.6.1.2  *t- ............................................................................................................. 102  3.6.1.3  *c-............................................................................................................. 102  3.6.1.4  *k- ............................................................................................................ 103  3.6.1.5  *q- ............................................................................................................ 104  3.6.2  Implosive/glottalized ................................................................................... 105  3.6.2.1  *ɓ- ............................................................................................................ 106  3.6.2.2  *ɗ- ............................................................................................................ 107  3.6.2.3  *ˀj- ............................................................................................................ 108  3.6.2.4  *ʔ-............................................................................................................. 109  3.6.3  Voiced stops................................................................................................. 110  3.6.3.1  *b- ............................................................................................................ 110  3.6.3.2  *d- ............................................................................................................ 111  3.6.3.3  *ɟ- ............................................................................................................. 112  3.6.3.4  *g- ............................................................................................................ 113  3.6.3.5  *ɢ-............................................................................................................. 114  3.6.4  Voiceless fricatives ...................................................................................... 115  3.6.4.1  *s-............................................................................................................. 115  3.6.4.2  *ɕ-............................................................................................................. 116  3.6.4.3  *x- ............................................................................................................ 117  3.6.4.4  *χ-............................................................................................................. 118  3.6.4.5  *h- ............................................................................................................ 118  3.6.5  Voiced fricatives .......................................................................................... 119  3.6.5.1  *z-............................................................................................................. 119  3.6.5.2  *ʑ- ............................................................................................................ 120  3.6.5.3  *ɣ- ............................................................................................................ 121  3.6.6  Voiced nasals ............................................................................................... 122  3.6.6.1  *m- ........................................................................................................... 122  3.6.6.2  *n- ............................................................................................................ 123  3.6.6.3  *ɲ- ............................................................................................................ 124  3.6.6.4  *ŋ- ............................................................................................................ 124  3.6.7  Voiceless nasals ........................................................................................... 125  3.6.7.1  *ʰm- .......................................................................................................... 126  3.6.7.2  *ʰn- ........................................................................................................... 126  3.6.7.3  *ʰɲ- ........................................................................................................... 127  ix 3.6.7.4  *ʰŋ- ........................................................................................................... 127  3.6.8  Voiced liquids and glides............................................................................. 129  3.6.8.1  *r- ............................................................................................................. 129  3.6.8.2  *l- ............................................................................................................. 130  3.6.8.3  *w-............................................................................................................ 131  3.6.9  Voiceless liquids and glides........................................................................ 133  3.6.9.1  *ʰr-............................................................................................................ 133  3.6.9.2  *ʰl- ............................................................................................................ 134  3.6.9.3  *ʰw- .......................................................................................................... 134  3.7  Summary .......................................................................................................... 136  Chapter 4................................................................................................................... 137  4.1  Introduction ..................................................................................................... 137  4.2  Tautosyllabic clusters ...................................................................................... 138  4.2.1  Clusters with medial *-r- ............................................................................. 140  4.2.1.1  *pr- ........................................................................................................... 140  4.2.1.2  *tr- ............................................................................................................ 141  4.2.1.3  *cr- ........................................................................................................... 142  4.2.1.4  *kr- ........................................................................................................... 143  4.2.1.5  *qr- ........................................................................................................... 144  4.2.1.6  *br- ........................................................................................................... 145  4.2.1.7  *gr- ........................................................................................................... 146  4.2.2  Clusters with medial *-l-.............................................................................. 147  4.2.2.1  *pl- ........................................................................................................... 147  4.2.2.2  *kl- ........................................................................................................... 148  4.2.2.3  *ɓl- ........................................................................................................... 149  4.2.2.4  *gl- ........................................................................................................... 150  4.2.3  Clusters with medial *-w-............................................................................ 151  4.2.3.1  Velar + *-w- ............................................................................................. 151  4.2.3.2  Uvular + *-w-........................................................................................... 152  4.2.3.3  Labial + *-w-............................................................................................ 153  4.2.3.4  Coronal + *-w- ......................................................................................... 155  4.2.3.5  Three-consonant clusters *CCw- ............................................................. 157  4.3  Speculative *-w-.............................................................................................. 158  4.4  Sesquisyllabic clusters..................................................................................... 162  4.4.1  Voiceless stop + voiceless stop.................................................................... 162  4.4.2  Voiceless obstruent + voiced stop ............................................................... 165  4.4.3  Voiced obstruent + voiceless stop ............................................................... 167  4.4.4  Voiceless stops + liquids/glides................................................................... 169  4.4.5  Voiced consonant + liquid/glide .................................................................. 172  4.4.6  Clusters with non-initial nasals.................................................................... 175  4.4.7  Other clusters ............................................................................................... 177  4.5  Reduction of sesquisyllabic clusters................................................................ 179  4.5.1  Processes of cluster adjustment ................................................................... 181  4.5.2  Monosyllabification of sesquisyllables........................................................ 185  4.6  Summary.......................................................................................................... 189  x Chapter 5................................................................................................................... 190  5.1  Introduction ..................................................................................................... 190  5.2  Inventory of PT rimes...................................................................................... 191  5.3  PT vowels: quality and quantity ...................................................................... 194  5.4  Solving Gedney’s Puzzle................................................................................. 198  5.5  Enriching the codas ......................................................................................... 208  5.6  Reconstructions of PT rimes ........................................................................... 213  5.6.1  Rimes with high vowels............................................................................... 213  5.6.1.1  Rimes with *i and *i: ............................................................................... 214  5.6.1.2  Rimes with *ɯ and *ɯ: ........................................................................... 216  5.6.1.3  Rimes with *u and *u:.............................................................................. 220  5.6.2  Rimes with mid vowels................................................................................ 222  5.6.2.1  Rimes with *e and *e: .............................................................................. 223  5.6.2.2  Rimes with *ɤ and *ɤ: .............................................................................. 226  5.6.2.3  Rimes with *o and *o:.............................................................................. 228  5.6.3  Rimes with low vowel *a and *a: ................................................................ 231  5.6.4  Rimes with diphthongs ................................................................................ 234  5.7  Summary.......................................................................................................... 236  Chapter 6................................................................................................................... 238  6.1  Introduction ..................................................................................................... 238  6.2  3+1: The PT tonal system................................................................................ 239  6.3  Directionality of tone change .......................................................................... 254  6.3.1  Segment-tone interaction ............................................................................. 255  6.3.2  Contextual variation..................................................................................... 256  6.3.2.1  Contour reduction..................................................................................... 257  6.3.2.2  Peak sliding .............................................................................................. 257  6.3.3  Perceptual maximization.............................................................................. 258  6.4  Tonal reflexes of PT in daughter varieties ...................................................... 259  6.5  Reconstructing the contrasts............................................................................ 267  6.5.1  Comparing *ABCD in daughter languages ................................................. 267  6.5.2  Vowel duration ............................................................................................ 271  6.5.3  Pitch height .................................................................................................. 274  6.5.4  Pitch contour ................................................................................................ 275  6.5.5  Voice quality................................................................................................ 277  6.5.6  PT as a “pitch-plus voice quality” tone language ........................................ 282  6.6  From PT tones to modern reflexes .................................................................. 282  6.7  Summary.......................................................................................................... 285  Chapter 7................................................................................................................... 286  7.1  Introduction ..................................................................................................... 286  7.2  Comparison of PT reconstructions .................................................................. 286  7.3  Tai subgrouping............................................................................................... 297  7.4  PT phonology in Kra-Dai context ................................................................... 304  7.5  Convergence of Tai varieties........................................................................... 308  7.6  Conclusion....................................................................................................... 311  xi Appendix A................................................................................................................ 313  Appendix B................................................................................................................ 323  Appendix C................................................................................................................ 378  References……………………………………………………….....………….…….391 xii LIST OF FIGURES Figure 1-1 Figure 1-2 Figure 1-3 Figure 2-1 Figure 2-2 Figure 2-3 Figure 3-1 Figure 3-2 Figure 4-1 Figure 4-2 Figure 4-3 Figure 4-4 Figure 4-5 Figure 4-6 Figure 4-7 Figure 4-8 Figure 5-1 Figure 6-1 Figure 6-2 Figure 7-1 Figure 7-2 The Kra-Dai family tree according to Edmondson and Solnit (Edmondson & Solnit 1997: 2) ................................................................ 5 Proposals for subgroup structure of Tai ................................................... 6 Tai dialects included in this study (see Appendix A)............................. 19 Sonority scale (based on Clements 1990: 292) ...................................... 34 Foot typology according to Hayes (1995) .............................................. 38 Phonological representation of PT prosodic word ................................. 65 Development of dorsal onsets in White Thai, Kapong, and Yay ........... 75 Development of dorsal onsets in White Thai, Debao, and Yay ............. 77 Weakening of *p.t- and * C̥ .d- ............................................................. 182 Implosivization of *t.n- and *C̥ .d- ....................................................... 183 Metathesis of *r.t-, and *p.q-................................................................ 184 Assimilation of *p.t-............................................................................. 184 Simplification of *t.h- .......................................................................... 185 Monosyllabification of *k.r- and *p.r- ................................................. 186 Monosyllabification of *k.t- and *t.h-.................................................. 187 Monosyllabification of *p.t-, *C̥ .d-, and *C̬ .t- ..................................... 188 Proposed solution for “Gedney’s Puzzle” ............................................ 206 Binary register split and neutralization of phonation type ................... 248 Schematization of the mechanisms of tone change .............................. 256 Tai subgrouping according to Li (1960)............................................... 297 Tentative subgrouping of Tai dialects .................................................. 298 xiii LIST OF TABLES Table 1-1  Table 1-2  Table 1-3  Table 1-4  Table 1-5  Table 1-6  Table 2-1:  Table 2-2:  Table 2-3:  Table 2-4:  Table 2-5:  Table 2-6  Table 2-7:  Table 2-8  Table 2-9:  Table 2-10  Table 3-1  Table 3-2  Table 3-3  Table 3-4  Table 3-5  Table 3-6  Table 3-7  Table 3-8  Table 3-9  Table 3-10  Table 3-11  Table 3-12  Table 3-13  Table 3-14  Table 3-15  Table 3-16  Table 3-17  Table 3-18  Table 3-19  Table 3-20  PT consonant inventory according to Li (1977)....................................... 8  PT consonant clusters according to Li (1977) .......................................... 9  PT inventory of simple vowels according to Li (1977).......................... 10  PT inventory diphthongs according to Li (1977) ................................... 11  PT rimes with final glides according to Li (1977) ................................. 12  Chart showing the laryngeal features of the initials at the time of tone split and the original PT tone categories................................................ 25  Reconstruction of certain etyma under different assumptions about the PT prosodic word ................................................................................... 43  Correspondences involving PT *ɗ- ....................................................... 44  Correspondences with *q- ..................................................................... 45  Correspondences involving *br- and *ɟ- ............................................... 47  Correspondences involving labial clusters ............................................ 48  Permissible onsets in Dehong Tai (based on Luo 1999: xiii)................. 54  Complex onsets in Nhaheun (edited from Ferlus 1998: ii) .................... 55  Complex onsets in Northern Roglai ....................................................... 58  Onsets and syllables in Kammu (from Pittayaporn 2005c).................... 61  Canonical shape of PT prosodic word................................................... 64  PT consonant inventory.......................................................................... 70  Voiceless dorsal correspondence sets..................................................... 73  Voiced dorsal correspondence sets......................................................... 76  Examples of voicing alternation ............................................................. 78  Voicing alternation items that show additional irregularities................. 80  Voicing alternation items that have clear Chinese origins ..................... 82  Voicing alternation items with possible Chinese origins ....................... 84  Distribution patterns regarding original voicing of some voicingalternation items..................................................................................... 85  Voicing alternation items that may have been part of PT their Kra-Dai cognates.................................................................................................. 87  Forms with Li’s simple aspirated onsets not found in NT ..................... 90  Forms with Li’s simple aspirated onsets that are found in all the three groups..................................................................................................... 92  PT clusters with medial */-r-/ ................................................................. 94  Examples of PT sesquisyllabic clusters that are reflected with aspiration ................................................................................................................ 96  Clusters consisting of a stop plus a liquid in PSWT .............................. 98  Schematization of the development of PSWT stop system from PT ..... 99  Forms showing aspirated initials in Chinese but unaspirated in Tai. ... 100  Etyma with PT *p-................................................................................ 102  Etyma with PT *t- ................................................................................ 102  Etyma with PT *c-................................................................................ 103  Etyma with PT *k-................................................................................ 104  xiv Table 3-21  Table 3-22  Table 3-23  Table 3-24  Table 3-25  Table 3-26  Table 3-27  Table 3-28  Table 3-29  Table 3-30  Table 3-31  Table 3-32  Table 3-33  Table 3-34  Table 3-35  Table 3-36  Table 3-37  Table 3-38  Table 3-39  Table 3-40  Table 3-41  Table 3-42  Table 3-43  Table 3-44  Table 3-45  Table 3-46  Table 3-47  Table 3-48  Table 3-49  Table 3-50  Table 3-51  Table 3-52  Table 3-53  Table 4-1  Table 4-2  Table 4-3  Table 4-4  Table 4-5  Table 4-6  Table 4-7  Table 4-8  Table 4-9  Table 4-10  Table 4-11  Table 4-12  Etyma with PT *q-................................................................................ 105  Etyma with PT *ɓ-................................................................................ 106  Etyma with PT *ɗ-................................................................................ 107  Etyma with PT *ˀj- ............................................................................... 108  Some etyma that Li reconstructs with *j-............................................. 109  Etyma with PT *ʔ-................................................................................ 110  Etyma with PT *b-................................................................................ 111  Etyma with PT *d-................................................................................ 112  Etyma with PT *ɟ- ................................................................................ 113  Etyma with PT *g-................................................................................ 114  Etyma with PT *ɢ-................................................................................ 115  Etyma with PT *s- ................................................................................ 116  Etyma with PT *x-................................................................................ 117  Etyma with PT *χ-................................................................................ 118  Etyma with PT *h-................................................................................ 119  Etyma with PT *z-................................................................................ 120  Etyma with PT *ʑ-................................................................................ 121  Etyma with PT *ɣ-................................................................................ 122  Etyma with PT *m- .............................................................................. 123  Etyma with PT *n-................................................................................ 123  Etyma with PT *ɲ-................................................................................ 124  Etyma with PT *ŋ-................................................................................ 125  Etyma with PT *ʰm- ............................................................................. 126  Etyma with PT *ʰn- .............................................................................. 127  Etyma with PT *ʰɲ- .............................................................................. 127  Etyma with PT *ʰŋ- .............................................................................. 128  Etyma with PT *r- ................................................................................ 130  Etyma with PT *l- ................................................................................ 130  Etyma with PT *w-............................................................................... 132  Some etyma that Li reconstructs with *w- ........................................... 132  Etyma with PT *ʰr- ............................................................................... 134  Etyma with PT *ʰl- ............................................................................... 134  Etyma with PT *ʰw- ............................................................................. 135  CC- clusters in PT ................................................................................ 139  Etyma with PT *pr- .............................................................................. 141  Etyma with PT *tr- ............................................................................... 142  Etyma with PT *cr- .............................................................................. 143  Etyma with PT *kr- .............................................................................. 144  Etyma with PT *qr- .............................................................................. 145  Etyma with PT *br- .............................................................................. 146  Etyma with PT *gr- .............................................................................. 146  Etyma with PT *pl- .............................................................................. 148  Etyma with PT *kl- .............................................................................. 149  Etyma with PT *ɓl- .............................................................................. 150  Etyma with PT *kw-, *xw-, *ɣw- and *ŋw- respectively .................... 152  xv Table 4-13  Table 4-14  Table 4-15  Table 4-16  Table 4-17  Table 4-18  Table 4-19  Table 4-20  Table 4-21  Table 4-22  Table 4-23  Table 4-24  Table 4-25  Table 4-26  Table 4-27  Table 4-28  Table 4-29  Table 4-30  Table 4-31  Table 4-32  Table 4-33  Table 4-34  Table 4-35  Table 4-36  Table 4-37  Table 5-1  Table 5-2  Table 5-3  Table 5-4  Table 5-5  Table 5-6  Table 5-7  Table 5-8  Table 5-9  Table 5-10  Table 5-11  Table 5-12  Table 5-13  Table 5-14  Table 5-15  Table 5-16  Table 5-17  Etyma with PT *qw-, and *ɢw- respectively........................................ 153  Etyma with PT *mw-............................................................................ 154  Etyma with PT *bw-, *pw-, and *ʰmw- ............................................... 155  Etyma with PT *rw- ............................................................................. 156  Etyma with PT *tw-, *nw-, *ʰrw-, and *ʰlw- respectively................... 157  Etyma with PT *krw-, *klw-, *qrw-, *trw-, and *crw- respectively ... 158  Etyma with PT *wu:............................................................................. 159  Etyma with PT *-wi:- ........................................................................... 160  Etyma with PT *we:, *wɤ: and *wo:.................................................... 161  Etyma with PT *p.t- ............................................................................. 163  Etyma with PT *k.t- ............................................................................. 164  Etyma with PT *p.q- and *q.p- ............................................................ 164  Etyma with PT *C̥ .b-............................................................................ 165  Etyma with PT *C̥ .d-............................................................................ 166  Etyma with PT *C̬ .t-............................................................................. 168  Etyma with PT *C̬ .k-, and *C̬ .q- respectively...................................... 168  Etyma with PT *k.r- ............................................................................. 169  Etyma with PT *p.r- ............................................................................. 170  Etyma with PT *C̥ .w- ........................................................................... 172  Etyma with PT *m.l- ............................................................................ 173  Etyma with PT *C̬ .r-............................................................................. 173  Etyma with PT *C̬ .l-............................................................................. 174  Etyma with PT *t.n- ............................................................................. 176  Etyma with PT *C̬ .n-............................................................................ 177  Monosyllabification in Tsat (Sagart 1993)........................................... 180  Inventory of PT simple vowels............................................................. 192  Inventory of PT diphthongs.................................................................. 192  Inventory of PT codas .......................................................................... 193  PT vowel system according to Li (1977) and Sarawit (1973).............. 195  Distribution of PSWT non-high vowels ............................................... 197  Items that unequivocally go back to PT */i:/, */ɯ:/, and */u:/, i.e. the ‘year’ set............................................................................................... 199  Items that unequivocally go back to PT */aj/, */aɰ/, and */aw/, i.e. the ‘egg’ set................................................................................................ 200  The ‘fire’ set of Gedney’s Puzzle correspondences ............................. 201  The ‘long’ set of Gedney’s Puzzle correspondences............................ 202  Comparison of solutions to “Gedney’s Puzzle” ................................... 203  Types of languages allowed by Li’s and Sarawit’s account................. 204  Tai varieties that do not belong to type 2 or type 3 .............................. 205  Etyma with irregular reflexes usually discussed under “Gedney’s Puzzle” ................................................................................................. 207  Shared etyma with final /-l/ in Saek and Kra ....................................... 209  Etyma with PT */en/ and */-el/............................................................. 210  Etyma with final */-c/........................................................................... 211  Final *-c in Tai and Be ......................................................................... 213  xvi Table 5-18  Table 5-19  Table 5-20  Table 5-21  Table 5-22  Table 5-23  Table 5-24  Table 5-25  Table 5-26  Table 5-27  Table 5-28  Table 5-29  Table 5-30  Table 5-31  Table 5-32  Table 5-33  Table 5-34  Table 5-35  Table 5-36  Table 5-37  Table 5-38  Table 5-39  Table 5-40  Table 5-41  Table 5-42  Table 5-43  Table 6-1  Table 6-2  Table 6-3  Table 6-4  Table 6-5  Table 6-6  Table 6-7  Table 6-8  Table 6-9  Table 6-10  Table 6-11  Table 6-12  Table 6-13  Table 6-14  Table 6-15  Table 6-16  Table 6-17  Table 6-18  Etyma with PT *i:#............................................................................... 214  Etyma with PT *iC and *i:C ................................................................ 215  Etyma with PT *ɯ:#............................................................................. 216  Etyma with PT *ɯn, *ɯ:n, and *ɯ:l respectively ............................... 217  Etyma with PT *ɯm, *ɯ:m, *ɯŋ, and *ɯ:ŋ respectively ................... 218  Etyma with PT *ɯ:p,*ɯ:t, and *ɯ:k respectively ............................... 219  Etyma with PT *wɯ:#.......................................................................... 220  Etyma with PT *u:#.............................................................................. 220  Etyma with PT *uC and *u:C respectively .......................................... 221  Etyma with PT *uk............................................................................... 222  Etyma with PT *e:# .............................................................................. 223  Etyma with PT *eC .............................................................................. 224  Etyma with PT *et, *en, and *el respectively ...................................... 225  Etyma with PT *e:C ............................................................................. 225  Etyma with PT *ɤC .............................................................................. 226  Etyma with PT *ɤ:C ............................................................................. 227  Etyma with PT *ɤ:# and *ɤ:l ................................................................ 228  Etyma with PT *o:#.............................................................................. 229  Etyma with PT *oC .............................................................................. 230  Etyma with PT *ok, and *oŋ respectively............................................ 230  Etyma with PT *o:C ............................................................................. 231  Etyma with PT *a:# .............................................................................. 232  Etyma with PT *aC and *a:C respectively........................................... 233  Etyma with PT *iə, *ɯə, and *uə respectively .................................... 234  Etyma with PT *iəw ............................................................................. 235  Etyma with PT *aɰ .............................................................................. 236  Correspondences between Tai and Chinese tones................................ 240  Phonological factors commonly found to condition tonal splits and mergers (modified from Gedney 1972a).............................................. 243  Pattern of tonal splits and mergers in White Tai .................................. 244  Pattern of tonal splits and mergers in Western Nung ........................... 244  Pattern of tonal splits and mergers in Southern Shan........................... 245  Pattern of tonal splits and mergers in Southern Thai ........................... 246  Pattern of tonal splits and mergers in Po-ai.......................................... 246  Pattern of tonal splits and mergers in Cao Bang .................................. 247  Western Nung tonal inventory.............................................................. 250  Pattern of tonal splits and mergers in Cao Bang .................................. 260  Pattern of tonal splits and mergers in Leiping...................................... 261  Pattern of tonal splits and mergers in Siamese ..................................... 262  Pattern of tonal splits and mergers in Black Tai................................... 263  Pattern of tonal splits and mergers in Wuming .................................... 264  Pattern of tonal splits and mergers in Liujiang..................................... 265  Pattern of tonal splits and mergers in Lianshan.................................... 266  PT *A in individual varieties................................................................ 268  PT *B in individual varieties................................................................ 268  xvii Table 6-19  Table 6-20  Table 6-21  Table 6-22  Table 6-23  Table 7-1  Table 7-2  Table 7-3  Table 7-4  Table 7-5  Table 7-6  PT *C in individual varieties................................................................ 269  Phonetic characteristics of PT tones..................................................... 271  Connection of *D to *B and *C ........................................................... 272  Siamese tones correlated with average vowel durations in msec (Abramson 1962: 107) ......................................................................... 273  Pattern of tonal splits and mergers in Nung Fan Slihng....................... 280  PT consonants and their correspondents in previous reconstructions .. 288  PT consonants and their correspondents in previous reconstructions .. 290  Examples of PT sesquisyllabic clusters and their correspondents in previous reconstructions ...................................................................... 291  PT rimes and their correspondents in previous reconstructions ........... 293  Phonological criteria for subgrouping of Tai languages ...................... 299  Proto-Tai and Sui reflexes of PKD back unrounded vowel. ................ 306  xviii LIST OF ABBREVIATIONS CT = Central Tai LC = Lungchow LMC = Late Middle Chinese MC = (Early) Middle Chinese NT = Northern Tai O = obstruent OC = Old Chinese PAN = Proto-Austronesian PIE = Proto-Indo-European PKD = Proto-Kra-Dai PSWT = Proto-Southwestern Tai PT = Proto-Tai S = sonorant SI = Siamese SEA = Southeast Asia SSP = Sonority Sequencing Principle SWT = Southwestern Tai Y = Yay -f = showing irregularity in the coda -i = showing irregularity in the onset -t = showing irregularity in the tone -v = showing irregularity in the vowel xix CHAPTER 1 BACKGROUND 1.1 Introduction The hardest task in comparative linguistics is to distinguish between types of similarities: universal tendencies, chance, diffusion, genetic retention, parallel development (Aikhenvald & Dixon 2001). Only when these different possible types of similarities are considered can we understand the role of common ancestry. The magnitude of this difficulty is reflected in the view that a similarity can be explained as shared inheritance only when other possibilities have been exhausted (Dench 2001). Successful reconstructions of proto-languages must therefore overcome this challenge. The reconstruction of Proto-Tai (PT), the ancestor of the Tai languages of Mainland Southeast Asia (SEA) is not an exception. Modern Tai languages share a great deal of structural similarities and identifiable phonological innovations. Yet, trying to reconstruct the phonology of PT has proven more challenging than it first seems. This is precisely because Tai languages have continually undergone the different waves of convergent trends that swept through the Mainland Southeast Asian linguistic area. In particular, various researchers have put forward their own reconstructed phonology of PT but these proposals disagree on major issues due in part to assumptions that they hold, be it explicitly or implicitly. These assumptions arose precisely from different interpretations of shared similarities among modern Tai languages by different researchers. Unfortunately, some of these assumptions have become axioms in the study of the history of Tai languages, and have thus kept those issues unsolved. With this in mind, this dissertation presents a new phonological reconstruction of PT constrained by explicit and motivated hypotheses. Essentially, it incorporates dialect data that have become increasingly available and discusses 1 explicitly the interpretation of these data. As suggested by the title, this dissertation takes as its task the reconstruction of the PT phonological system rather than detailed analyses of individual etyma, or a complete lexicon of PT. The widely accepted PT reconstruction is due to Li (1977). This reconstruction has been the basis of good scholarship on the history of the Tai languages but like other essays on PT reconstruction (Ferlus 1990; Haudricourt 1948; Liang & Zhang 1996; Luo 1997; Nishida 1954; Sarawit 1973), it lacks sufficient consideration of the issues mentioned above. This study goes beyond Li and other historical research on Tai languages in that it takes into serious consideration the structure and the phonotactics of the PT prosodic word. In particular, I claim that PT was a sesquisyllabic language that allowed both sesquisyllabic and monosyllabic prosodic words. The PT prosodic word consisted of three main components: the onset, the rime, and the tone. The rime is, in turn, subdivided into the vocalic nucleus and the coda. While I show that the PT vocalism was much simpler and symmetrical than conventionally reconstructed, I claim that the onset and the coda systems are richer than previously assumed. Furthermore, I propose a concrete albeit preliminary reconstruction of the PT tonal categories. 1.2 Tai languages Tai is a technical term that is used to refer to various related ethnolinguistic groups in SEA, although only some Tai-speaking groups call themselves “Tai” or its variants. Though pronounced identically in English, “Tai” is used to refer to the group as a whole or to any language belonging to the family, while “Thai” is preserved for the dominant language of Central Thailand, also known as “Siamese”. In SEA five language families have been identified: Sino-Tibetan, Austroasiatic, Austronesian, Hmong-Mien, and Kra-Dai (formerly known as Tai-Kadai). The Tai languages, a sub- 2 branch of Kra-Dai, will be the subject of this dissertation. The hypothetical parent language from which all Tai languages develop is called Proto-Tai (PT). It has been speculated that PT was spoken in the area around the Guangxi-Vietnam border areas (Diller 2000; Gedney 1995). Nowadays Tai languages are currently spoken in China, Vietnam, Laos, Cambodia, Thailand, Malaysia, Myanmar, and India. The best known members of this group are Thai, Lao, and Shan, of which the first two are the official languages of Thailand and Laos respectively. Some other better-known ethnonyms used to refer to Tai-speaking groups include Nung, Thu, Zhuang, Bouyei, Giay, and Saek (Edmondson & Solnit 1997). These Tai languages occupy the geographical center of the Southeast Asian mainland, and have been active adopters and spreaders of many areal traits with languages to their north, e.g. Chinese, Tibeto-Burman, and Hmong-Mien, as well as further south, e.g. Austroasiatic and Austronesian (Edmondson & Solnit 1988) Typologically, modern Tai languages share a number of phonological characteristics (Diller 2000; Gedney 1989b). All Tai languages are basically monosyllabic and tonal. A given modern language can best be described in terms of the distinction made between various parts of the syllables (Gedney 1989b: 13). Each syllable can be divided into onset, a vocalic nucleus, and an optional final consonant. With respect to the onset, most Tai languages have a fairly rich inventory which includes simple onsets and clusters of the stop-glide and/or stop-liquid types. As regards the vowel system, most languages make a three-way distinction in height (high, mid, and low) and a three-way distinction in tongue position and rounding (front, back rounded, and back unrounded). Some languages also show length contrast throughout their vowel systems, while others only contrast short /a/ and long /a:/. Final consonants are very limited in modern Tai languages. Most languages have stops and nasal finals in the labial, dental, and velar places of articulation. Some also have an 3 additional final glottal stop. Tones in Tai languages always show distinctions in both pitch height and contour. The number of tones in each language ranges from 4-7. In many better-studied varieties, the distributions of the tones are conditioned by the syllable structure; fewer tones can occur in checked syllables, i.e. syllables closed by an obsturent. However, as Diller (2000: 13) correctly points out, these characteristics are found widely throughout SEA. These Tai languages form a branch of the Kra-Dai1 language family, formerly known as Tai-Kadai. According to the subgroup structure proposed in Edmondson and Solnit (Edmondson & Solnit 1997: 2) shown in Figure 1-12, the closest relative of Tai according to this classification is the Be language (临高 Lín’gāo) of Hainan. The majority of Kra-Dai languages outside of Tai are spoken in the extreme South of China. The Kam-Sui group comprises Kam (侗 Dòng), Sui (水 Shǔi), Maonan (毛南 Máonán), Mulam (仫佬 Mùlǎo), Ai-cham (锦 Jǐn), Then (佯僙 Yánghuáng) and Mak (莫 Mò). Closely related to Kam-Sui is Lakkja (拉珈 Lājiā) and Biao (标 Biǎo), which form their own subgroups. The Hlai group (黎 Lǐ) consists of closely related dialects spoken by indigenous population of Hainan Island. Lastly, the Kra group consists of languages spoken by smaller populations, including Gelao (仡佬 Gēlǎo), Lachi (拉基 Lājī), Buyang (布央 Bùyāng), Pubiao (普标 Pǔbiǎo), Yerong (耶容 Yēróng), and Laha. These smaller languages are scattered in Guizhou, Yunnan, and Guangxi Provinces of China, except for Laha which is spoken in Vietnam. 1 This label is due to Ostapirat (2000)’s seminal work on the reconstruction of Proto-Kra. 2 See Ostapirat (2005: 108) for another possible subgroup structure. 4 Figure 1-1 The Kra-Dai family tree according to Edmondson and Solnit (Edmondson & Solnit 1997: 2) Although membership of Tai in the Kra-Dai family is not-controversial, the relationship of Tai to other language families is still open for investigation. Kra-Dai has been grouped with all the other families in SEA, including Hmong-Mien (Kosaka 2002) and Austroasiatic (de Lacouperie 1886), but the most widely accepted views used to be that Tai belongs with Chinese in the Sino-Tibetan family (e.g. Conrady 1896; Grierson 1903; Li 1976; Luo 1997; Manomaivibool 1975; Schmidt 1926; Wulff 1934). However, the “Sino-Tai” view is becoming superseded by the competing view that links Kra-Dai with Austronesian (Benedict 1942, 1975, 1997b; Ostapirat 2005; Sagart 2004, 2005; Schlegel 1902; Wulff 1942). This view is best represented by the “Austro-Tai” hypothesis proposed by Benedict (1942; 1975), who believes that KraDai, Austronesian, and Hmong-Mien form one linguistic stock. Although Benedict’s linguistic evidence and methodology has been severely criticized, his proposal to connect Kra-Dai with Austronesian merits close attention and has found support in more recent works by Ostapirat (2005) and Sagart (2004; 2005). Although this study 5 does not directly address the issue of the affinity of Kra-Dai, a non-trivial link between Tai and Austronesian will become progressively clearer all through the dissertation. Within the Tai branch, the conventional classification is that proposed by Li (1960; 1977). It divides Tai varieties into three primary subgroups: Southwestern Tai (SWT), Central Tai (CT), and Northern Tai (NT). The ancestral languages of SWT and NT have been proposed by various researchers (e.g. Brown 1965; Jonsson 1991; Kosaka 1992; Li 1977; Pittayaporn 2008b, to appear-b). However, other alternative subgrouping schemas have also been proposed (Chamberlain 1975; Haudricourt 1956; Luo 1997). Common among many of the alternative view is the recognition of similarities between Li’s SWT and CT as reflected by Haudricourt (1956)’s explicit claims that the two comprise a sister branches of NT 3 . Li’s and Haudricourt’s classification schemas are given in Figure 1-2. PT PT NT CT SWT Li (1977) NT CT SWT Haudricourt (1956) Figure 1-2 Proposals for subgroup structure of Tai Note that, despite his bipartite subgroup structure, Haudricourt (1956) claimed that a few understudied dialects occupy an intermediate position between SWT/CT and NT. The result of this dissertation will show quite clearly that among Li’s three 3 Haudricourt (1956) uses “Thai” to refer restrictively to the subgroup consisting of Li’s SWT and CT and “Dioi” to Li’s NT. 6 subgroups SWT is the only valid one (see §7.3). However, Li’s subgroup labels are useful and convenient labels in describing the distribution of linguistic traits shared by modern Tai varieties, and help relate the current research to the existing body of literature on Tai languages. Therefore, in this dissertation I use the terms SWT, CT, and NT as generic labels to refer grossly to clusters of languages that share a certain set of developments. In particular, I define the three clusters using the same criteria that Li (1960) uses in his proposal but without adopting his claim that they constitute genealogical subgroups of Tai. The only genealogical subgroup that I assume is SWT, which I have shown elsewhere to be a valid one (Pittayaporn 2008b). The issue of Tai subgrouping will be discussed in Chapter 7. 1.3 Conventional reconstruction of PT and pending issues Despite a number of competing reconstructions of PT, the most widely adopted is, beyond doubt, the system proposed by Li Fang-Kuei, especially in his Handbook of Comparative Tai (Li 1977). Not only does this monumental work embody the advances in the study of Tai languages from the 40’s through the 70’s, it has also been a basis for Tai linguistics from its first appearance to the present. Even though alternative reconstructions do exist (Ferlus 1990; Haudricourt 1948; Liang & Zhang 1996; Luo 1997; Nishida 1954, 1955; Sarawit 1973), it is not amiss to refer to Li’s construction as the conventional reconstruction of PT. In this section, I first outline the conventional reconstruction of PT and then proceed to discuss open issues in the reconstruction of PT phonology. 1.3.1 Li’s reconstruction The conventional reconstruction assumes that the development of PT etyma can be accounted for in terms of single syllables consisting of the onset, the vocalic 7 nucleus, the optional coda, and the tone (Edmondson & Solnit 1997; Li 1977). In this system, the onset position shows a four-way phonation-type contrast among consonants: 1) aspirated voiceless, 2) plain voiceless, 3) glottalized, and 4) plain voiced. However the phonation-type contrast is neutralized in coda position so that only plain voiceless stops and voiced nasals and glides are allowed. Table 1-1 gives the inventory of consonants of PT as conventionally reconstructed. Those sounds that can occur in the coda position are underlined. Table 1-1 PT consonant inventory according to Li (1977)4 stops fricatives nasals liquids and glides aspirated voiceless glottalized voiced voiceless voiced voiceless voiced voiceless voiced labial *pʰ *p *ˀb *b *ʰm *m *ʰw *w alveolar *tʰ *t *ˀd *d *s *z *ʰn *n *ʰr *ʰl *r *l palatal *cʰ *c *ɟ *ʰɲ *ɲ *ˀj *j velar *kʰ *k *g *x *ɣ (*ʰŋ) *ŋ glottal *ʔ *h These consonants can be combined into clusters of two consonants. While the initial consonants may be a stop, a fricative, or a nasal, the medial must either be a liquid or a glide. Note that Li (1977: 236-242) posits labiovelar units *kw-, *kʰw-, *gw-, *ŋw-, *xw-, and *ɣw- but states that it is not clear whether they are unitary phonemes or clusters consisting of a velar consonant followed by medial *-w-. 4 Li’s segments are re-transcribed as discussed in §1.4.3. 8 Table 1-2 PT consonant clusters according to Li (1977) aspirated stops voiceless stops glottalized voiced stops voiceless fricatives voiced fricatives nasals -r-l-w-r-l-w-r-l-w-r-l-w-r-l-w-r-l-w-r-l-w- labial *pʰl/r alveolar *tʰr *tʰl palatal *pr *tr *pl *tl *ˀbl/r *ˀdl/r *br dl *bl dr *fr- *vr *vl *ml/r *mw *nl/r velar *kʰr *kʰl *kʰw *kr *kl *kw *gr *gl *gw *xw *ɣw *ŋw glottal In addition to clusters with medial liquids shown above, there are also clusters with medial glides. Li’s system has three glides altogether: *j, *w, and *ɰ. While *j, and *w can occur as onset, medial glide, or coda, *ɰ is allowed only medially and finally. This set of medial glides requires special consideration. In Handbook of Comparative Tai, Li (1977) posits only two medial consonants explicitly: *-l- and *-r-. However, the three medial glides can be uncovered in different parts of the system. The first and the main part is the vocalism. Li (1977: 298) distinguishes two 9 types of high vowel in diphthongs—accented *i, *ɯ, and *u on one hand, and unaccented *i̯ , *ï̯ , and *u̯ on the other. For example, he reconstructs a contrast between *ie (1977: 280-281) vs. *i̯ e (Li 1977: 269-270). Since such a fine contrast seems implausible, I re-interpret Li’s unaccented *i̯ , *ɯ̯ , and *u̯ as glides *j, *ɰ, and *w respectively. According to this reinterpretation, the contrast between *ie and *i̯ e is in fact a contrast between a diphthong *ie and a sequence *je. Therefore, under this interpretation the conventional reconstruction allows for three medial glides *-j-, *-ɰ-, and *-w-, in addition to the two medial liquids *-l- and *-r.- As for the rime, the conventional reconstruction of PT has nine simple vowels without length contrast, and a large number of complex vowels. These vowels can be optionally followed by a final consonant from the set of permissible codas shown in Table 1-1 (underlined). Among simple vowels, distinctions are made for three vowel heights, backness, and rounding as shown in Table 1-3. Table 1-3 PT inventory of simple vowels according to Li (1977) front back unrounded unrounded rounded high *i *ɯ *u mid *e *ɤ *o low *ɛ *a *ɔ These simple vowels can be combined together to form diphthongs. One of the most prominent properties of Li’s inventory of complex vowels is the contrast between diphthongs with accented high vowels and those with unaccented high vowels (Li 1977: 298). However, I consider only the former as true diphthongs in the 10 conventional reconstruction. This is because the latter is re-interpreted as a sequence of a glide followed by a vowel. For comparison, these sequences are given in parentheses in Table 1-4. Table 1-4 PT inventory diphthongs according to Li (1977) second first *i high *ɯ *u mid low *e *ɤ *o *ɛ *a *ɔ *i high *ɯ *u *ie *iɤ (*je) (jo) (jɛ) *uɯ *ɯu *ɯe *ɯɤ ɯo ɯɛ ɯa ɯɔ (*ɰu) (ɰa) (ɰɔ) *ue *uɤ uo ua uɔ (*wɤ) (wo) (wa) (wɔ) In addition to the diphthongs shown in Table 1-4, more diphthongs and triphthongs are included in Li’s reconstruction. However, these complex vowels only end in *i, *ɯ, or *u. These final high vowels can be re-interpreted as post-vocalic *-j, *-ɰ, and *-w respectively, parallel to Li’s unaccented high vowels in the prevocalic position discussed above. Table 1-5 presents the inventory of these re-interpreted rimes with final glides in the conventional reconstruction. 11 Table 1-5 PT rimes with final glides according to Li (1977) *-j *-ɰ *-w high *i *iw *ɯ *ɯj *u mid *e *ej *eɰ *ew *ɤ *ɤj *wɤj *ɤɰ *ɤw *jɤw *o *oɰ *ow low *ɛ *ɛj *ɛw *a *aj *ɰaj *aw *ɔ *ɔj diphthongs *iɤ *iɤj *ia *iaj *uɤ *uɤj *ua *uaj In addition to the onset and the rime, each PT word in the conventional reconstruction also has tone *A, *B, *C, or *D. While *A, *B, and *C occurred in syllables ending in a vowel or a nasal, or ‘non-checked’ syllables, *D only occurred in syllables with a final obstruents, or “checked syllables.” Li (1977: 25) states that *D only occurred in checked syllables, in which tonal contrast was neutralized, However, he posits *D as a separate tone because it is impossible to identify it with the one of the other three tones. Moreover, the conventional reconstruction does not provide concrete reconstruction of the abstract tonal categories *A, *B, *C, and *D. 12 1.3.2 Outstanding problems The conventional reconstruction of PT phonology outlined above has played a major role in the historical study of Tai languages but many questions still await solutions. These problems include the unaccounted onset correspondence sets, the complexity of the vocalic system, the peculiarities of the displaced NT language Saek, spoken in northeastern Thailand, and the abstractness of tones. In this sub-section, I briefly introduce the issues and then assess them from a holistic perspective. The first problem concerns onset correspondences. Although the conventional reconstruction nicely accounts for the majority of the correspondence sets, minor sets of data are still left as puzzles. The best-know of these sets is probably the so-called “voicing alternation,” which refers to an apparent discrepancy regarding voicing of the onsets in different dialects. This discrepancy is generally described as a disagreement between SWT and CT on one hand, and NT on the other. For a sizeable set of eytma, reflexes in SWT and CT point to earlier voiceless onsets while their NT counterparts indicate original voiced onsets (Gedney 1989a). Take the etyma ‘to arrive’ and ‘ripe, cooked’ as examples. Based on SWT and CT reflexes alone, they should be reconstructed with *tʰ and *s- respectively. In contrast, they should be reconstructed as *d- and *z- respectively if only NT reflexes are taken into consideration. The conventional reconstruction fails to account for this set of data. In addition to the “voicing alternation” just described, there are many other sets of onset correspondences that still await explanation. These data sets often involve a small subset of etyma but the correspondence seems too systematic to be disregarded as irregular. One example is *tr- which Li reconstructs for etyma like ‘eye’ and ‘to die’. This cluster is reflected as /t-/ in the SWT language Siamese and the NT dialect of Po-ai, but as /h-/ in the CT dialect of Lungchow. At first glance this choice of reconstruction seems adequate but data from other varieties cast doubt on this 13 treatment of the correspondence. In particular, the CT dialect of Bao Yen from my own fieldwork regularly has /pʰj-/ for this set of etyma, indicating that the original PT onset must have contained a labial segment. The conventional reconstruction has no way to account for this correspondence. The second problem for the conventional reconstruction is the complexity of the vowel system. As apparent in the review above, Li’s reconstructed PT vocalic system contains a massive set of complex vowels with very fine phonemic contrasts. Even after re-interpreting many of his complex vowels as sequence of vowels and glides, the problem still remains equally damaging. A large part of this problem has to do with the paradigm of data sometimes referred to as “Gedney’s Puzzle” (Gedney 1972b). This problematic paradigm of vowel correspondence consists of four sets of vowel correspondences involving high vowels. A number of Tai etyma unequivocally go back to PT simple vowels *i, *ɯ, and *u in the conventional reconstruction. Similarly, another set of etyma clearly go back to Li’s *ɤj, *ɤw, and *ɤɰ. However, in addition to these two sets, there are a number of items that seem to point to earlier *i, *u, and *ɯ in some languages but *ɤj, *ɤw, and *ɤɰ in others. To complicate the picture, the reverse patterns also exist. The massive vowel system in the conventional reconstruction is in large part an attempt to account for this paradigm of data. The third challenge for the conventional reconstruction of PT is the aberrancy of Saek, a displaced NT language spoken in northeast Thailand and central Laos. This language shows many peculiarities that cannot be reconciled within the conventional model of PT phonology (Gedney 1989c; Haudricourt 1963b). For example, there are two reflexes of the PT sound that Li reconstructs as *ˀb- in Saek. The first one is the expected /b-/ as in ‘leaf’ and ‘to fly’. The other reflex is /v-/ as in ‘shoulder’, and ‘thin’. This pattern is left unexplained in Li (1977) as there does not seem to have been any conditioning environment for the split of PT *ˀb-. Another example of Saek 14 peculiarities, is the famous /-l/, found nowhere else in Tai. While recognizing that Saek final -l might have been inherited from PT, Li does not include a final liquid in the reconstruction due to lack of data. The last but in no way less challenging issue for the conventional reconstruction is the abstractness of PT tones. Although it is well-established that PT had four tonal categories, it is not clear what they were like. In other words, it is unknown what the contrastive characteristics of *A, *B, and *C, were and how *D related to the other three categories. The conventional reconstruction remains silent about the phonological content of these categories. From a holistic perspective, these problems largely stem from one serious shortcoming. In particular, Li does not address explicitly the structure of the PT prosodic word and its role in the reconstruction. Since words serve as the basic domains for comparative reconstruction, the researcher’s assumptions about the phonotactics of the prosodic word in the PT inevitably predetermine the reconstruction. Li’s reconstruction assumes at least implicitly that PT was typologically very similar to its modern descendants. Among other things, it assumes that like all modern Tai languages PT was monosyllabic. It also assumes a priori that the PT vowel system did not have a quantity contrast. These are assumptions that cannot be taken for granted. Recent research (e.g. Benedict 1982; Brunelle 2008; Ferlus 1975, 1992; Thurgood 1996, 2005) has shown how language contact leads to restructuring of the canonical word shape, how monosyllabic tonal languages arise from languages that have non-tonal polysyllabic canonical roots, etc. For example, all modern Chinese varieties have fully-developed tone systems, but it is uncontroversial that their ancestor Old Chinese was non-tonal (Baxter 1992). Therefore, from a holistic perspective, the issue of the PT prosodic word must not be taken for granted in the search for a solution to these problems. 15 In addition, these problems result from lack of comparative data. Li’s Handbook of Comparative Tai (Li 1977) is an honorable piece of work because it brings together lexicographical data from various dialects available during those early decades. However, as Gedney pointed out in 1967, most major Tai-speaking areas were still understudied (Gedney 1989b). Only for Thailand and southern Guizhou were dialect studies on fine geographical grids available. As evidenced by the discovery of Saek, data from previously unknown dialects may lead us to question old assumptions and provide new insights into the reconstruction of PT phonology. 1.4 Methodology, data, and notation Given the challenges discussed above, it is imperative that assumptions about PT phonology be critically assessed. Moreover, it is of equal importance to minimize gaps in our knowledge of modern dialects. Therefore, this dissertation approaches the reconstruction of PT by using a large set of comparative data that have now become available as a basis for discussing the problematic aspects of the conventional reconstructions. In this section, I first describe the methodology and data used in this dissertation before proceeding to discuss the notation used. 1.4.1 Methodology The method employed in this dissertation is the Comparative Method which first developed in the 19th century to study the history of Indo-European languages. However, it emphasizes three important methodological facets that are sometimes taken for granted in the reconstruction of proto-languages: 1) use of external evidence, 2) role of diffusion, and 3) subgroup assumptions. With respect to use of external evidence, in this dissertation, cognates from Kra-Dai languages outside of Tai, shared vocabulary with other language families of 16 SEA, and typological considerations are used strictly as additional support for reconstructions proposed on the basis of Tai-internal evidence. For example, the reconstruction of final *-l (see §5.5) is motivated by the presence of final /-l/ in Saek and the vowel reflexes in Qinzhou. Final /-l/ preserved in cognates in the Kra language Laha only lends additional yet strong support for positing PT *-l. Similarly, PT *C̥ .d(see §2.3.2 and §4.4.2) was reconstructed solely on the basis of sound correspondences among modern Tai dialects. By citing the Proto-Austronesian form *qudip ‘alive’ and *tidəm ‘dark’, I merely use the non-trivial similarities as additional support. I do not use these Proto-Austronesian forms as evidence for the reconstruction of PT complex onsets. Nor do I claim that Tai is genetically related to Austronesian. The role of diffusion is also a factor that cannot be ignored in applying the Comparative Method. Certain traits that are shared widely among Tai dialects may in fact be innovations. A good example is the case of PT uvular consonants (see §3.3). Although no attested modern Tai dialects have a distinctive uvular series, cases of velar stop /k-/ in some dialects corresponding to /h-/ in others unequivocally indicate that PT must have had uvular consonants *q-, *χ-, and *ɢ-, which contrasted with velar *k-, *x-, *g-, and *ɣ-. Another example is PT final *-c (see §5.5), which is not attested in any modern Tai varieties. However, cases where Saek /-k/ corresponds to /-t/ in other varieties must go back to a final palatal *-c. This interpretation is supported by the Kra-Dai languages Be and Hlai. The lack of uvular series and final *-c must undoubtedly be explained as traits that diffused from one Tai dialect to another throughout the Tai-speaking area. These two cases are good illustrations of why one should not underestimate the effect of diffusion in applying the Comparative Method. 17 Furthermore, the interpretation of sound correspondences in this dissertation does not make references to any particular proposal of genealogical subgrouping of Tai languages. This is because arriving at a subgroup structure without establishing the proto-language is theoretically highly problematic. In particular, interpreting shared similarities as retentions or innovations depends entirely on the reconstructed PT language. As subgrouping arguments take shared innovations as evidence, it is not possible to assume subgroups of Tai languages without having a clear understanding of the proto-language. Although I have established elsewhere (Pittayaporn 2008b) that SWT forms a valid subgroup within Tai (and I will, for presentational purposes, occasionally refer to Proto-SWT), the status of SWT varieties as a subgroup is never used here as evidence for PT reconstruction. 1.4.2 Data and fieldwork The data used in this dissertation cover a broad geographical area. Figure 1-3 is a map of all dialects included in this study. The varieties included in this dissertation belong to all three dialect groups: SWT, CT, and NT. They represent most of the different geographical areas where Tai languages are spoken, including Thailand, Laos, Malaysia, Myanmar, northeast India, Yunnan, Guangxi, and Guizhou. The only two areas for which availability of data does not permit close examination include Central and Eastern parts of Laos, Central Vietnam, and the interior of northern Vietnam. However, reports on varieties from these areas (Chamberlain 1984; Đoàn 1996; Haudricourt 1960; L'Écôle Française d'Extême-Orient 1938; Simmonds 1965) have also been taken into account. 18 Figure 1-3 Tai dialects included in this study (see Appendix A) 19 The dialect data used in this dissertation comes from both published materials and my own fieldwork. Appendix A lists all the varieties included in this study and their sources. In total, data from 68 dialects are used in the reconstruction. 1.4.2.1 Published materials The main set of data used is from Gedney’s unpublished wordlist (Gedney n.d.), which has now been published as part of the William J. Gedney’s Comparative Tai Source Book (Hudak 2008). Another main source of lexical data comes from a comprehensive survey of Zhuang dialects of Guangxi and Yunnan 壮语方言研究 Zhuàngyǔ Fāngyán Yánjiū (Zhang et al. 1999). In addition to these two sources, other resources that play invaluable roles in this dissertation include published and unpublished wordlists, dictionaries, and dialect descriptions in English, Thai, and Chinese (see Appendix A). Worth special mention is the survey of Bouyei dialects 布 衣語調查报告 Bùyīyǔ Diàochá Bàogào (Chinese Academy of Sciences 1959) which includes 40 localities altogether. These Bouyei dialects will be referred by their numbers in the original source, e.g. Bouyei (point 4) or Bouyei(4) refers to dialect point 4 in Bùyīyǔ Diàochá Bàogào. 1.4.2.2 Fieldwork Equally important is my own fieldwork carried out on a number of dialects in Thailand, Vietnam, and China. Altogether I carried out fieldwork on five Tai varieties. Three of them are spoken in northern Vietnam, one of the most understudied Taispeaking areas. During the winter of 2008, I collected data from the Tai dialects of Sapa, Bao Yen, and Cao Bang. The first two are spoken in Lao Cai province while the third is spoken in Cao Bang province. In Thailand, I carried out fieldwork on the 20 Kapong dialect of Phu Thai spoken in Sakon Nakhorn Province in 2005. In China, I collected data of the Shangsi dialect in southeastern Guangxi in spring of 2009. Data from these five varieties are invaluable for the task of reconstructing of PT phonology. 1.4.2.3 Representative dialects Given the number of dialects included in this study, it is not possible to discuss and present all data in this dissertation. In the discussion, three dialects from the three clusters are consistently used: Siamese, Lungchow, and Yay. They represent SWT, CT, and NT respectively. These dialects are chosen mainly because of the richness and quality of the lexical data available. Moreover, choosing these dialects allows easy comparison between this dissertation and Li (1977). Siamese and Lungchow are used in Li (1977) as representative dialects. Yay is not included in Li (1977) but his NT representative Po-ai, is closely related to it. 1.4.2.4 Tai-external data In addition to data from Tai varieties, data from other Kra-Dai languages as well as languages from other families are also used as additional evidence. Although this dissertation frequently makes references to the Kra-Dai language Lakkja (LThongkum 1992; Liu 1999; Pittayaporn 2008a), Proto-Austronesian (Wolff in press), Late Middle Chinese (Pulleyblank 1991), (Early) Middle Chinese (Schuessler 2009), Later Han Chinese (Schuessler 2007, 2009) and Old Chinese (Schuessler 2007, 2009), the reconstruction is based solely on Tai-internal evidence. As discussed above, the Tai-external data are used only as clues, or to help confirm certain choices of reconstruction. Unless otherwise indicated, forms cited in this dissertation from these languages are from the above sources. Sino-Tai etyma are cited from Li (1976) and/or Manomaivibool (1975), except where indicated. 21 1.4.3 Notation Because this dissertation incorporates data from diverse sources and compares reconstructions by different researchers, it is necessary to establish notational conventions that facilitates comparisons between different Tai dialects and different studies on PT phonology. All modern dialect forms as well as reconstructed PT forms that appear in this dissertation have been re-transcribed according to the IPA standard (International Phonetic Association 1999) as detailed below. 1) Aspiration is transcribed as a superscripted [ʰ] rather than a normal [h], e.g. Siamese /pʰa:B1/ ‘to split’ rather than /pha:B1/, and PT *ʰma: ‘dog’ rather than *hma:. 2) Glottalization is transcribed as a superscripted [ˀ] rather than a normal [ʔ], e.g. Debao /ˀja:A1/ ‘medicine’ rather than /ʔja:A1/. 3) The palatal semivowel is transcribed as [j] rather than [y] which represents the high front rounded vowel, e.g. Siamese /ja:A1/ ‘medicine’ rather than /ya:A1/. 4) Syllable-final semivowels are treated as glides [j], [ɰ], and [w], e.g. Li’s *vɛiA ‘fire’, *mɛuA ‘cat’, and *hmoɯB ‘new’ are retranscibed as *vɛjA, *mɛwD, and *ʰmoɰB respectively. 5) Li’s unaccented high vowels *i̯ , *ï̯ , and *u̯ are re-transcribed as glides *j, *ɰ, and *w respectively, e.g. Li’s *pi̯ etD ‘duck’, *fï̯ aC ‘cloud’ and *dlu̯ omA ‘wind’ are written as *pjet D, *fɰaC, and *dlwomA respectively (see §1.3.1). The undermark diacritic indicates that Li considers these semivowels. 22 6) High and mid back unrounded vowels are transcribed as [ɯ] and [ɤ] rather than [ï] and [ə], e.g. Lungchow /mɯ:A2/ ‘hand’ rather than /mï:A2/, and /fɤnA2/ ‘firewood’ rather than /fənA2/. The symbol [ə] is reserved for reduced and neutral vowels only. 7) [tɕ] and [č] both represent the same voiceless palatal affricate but are considered stops phonologically and re-transcribed as [c], e.g. Siamese /comA1/ ‘to sink’ rather than /tɕomA1/ or /čomA1/. Similarly, [dʑ] and [ǰ] are re-transcribed as [ɟ]. In addition, [ɕ] is used instead of [š], e.g. /ɕamA1/ ‘to sink’ rather than /šamA1/. Two exceptions are Li’s *č- and *ǰ-, which are retranscribed as *tɕ- and *dʑ- because Li (1977: 164-173) explicitly states that the two PT phonemes were affricates. 8) Long vowels are indicated by [:] rather than double vowels, e.g. Siamese /ja:A1/ ‘medicine’ rather than /jaaA1/. The notation of tones in modern dialects deserves special attention. It is the area where the largest number of different practices co-exists. These conventions range from using diacritics, numbers, or tone letters, to using syllable-final letters (Edmondson & Solnit 1997). To allow for ease of analysis and presentation, notations containing etymological information are widely used in historical work. A few systems of this sort are available, but all share one common characteristic: they make reference to PT tone categories *A, *B, *C, and *D and the laryngeal features of the leftmost onset consonants. An essential method used in approaching the onset-tone interactions in the history of Tai languages is known as the ‘tone-box’ method, formalized by Gedney (1972a). It is well-established that PT had a system of three tones on non-obstruentfinal syllables, or “non-checked syllables”: *A, *B, and *C. There is no tonal contrast 23 on obstruent-final syllables, or “checked syllables”; this type of syllables are then said to be toneless, or conventionally to have *D tone. Some time after PT had diversified into various languages, sound changes involving splits in the tonal system conditioned by the laryngeal feature of the initial consonants took place and spread through out the SWT-speaking territory (Gedney 1991; Pittayaporn 2008b). Each of the four PT tones then split into two or more tones. Some of the resulting tones then merged or further split. Checked syllables usually underwent an additional split conditioned by vowel length. In talking about initials and tones in Tai languages, it is useful to draw a chart showing the laryngeal features of the initials at the time of tone split and the original PT tone categories. In Table 1-6, the order of the laryngeal features presented vertically, and the order of the PT tones, presented horizontally, must not be altered. A lexical item is then annotated with the name of the box that corresponds to its PT initials and tones. For example, Thai /ma:5/ ‘dog’ which comes from PT *ʰma:A is written as /ma:A1/ to signal that the original tone was the *A tone and the original initial was voiceless. Note that *DS and *DL represent PT *D on syllables with short and long vowels respectively. 24 Table 1-6 Chart showing the laryngeal features of the initials at the time of tone split and the original PT tone categories. *A *B *C *DS *DL aspirated *f-, ʰm-, pʰ-,… A1 B1 C1 DS1 DL1 1 unaspirated *p-,… A1 B1 C1 DS1 DL1 glottalized/implosive *ɓ-,… A1 B1 C1 DS1 DL1 2 voiced *v-, m-, b-,… A2 B2 C2 DS2 DL2 Although it is well established that the basic type of tone split is the so-called “register split”, which is conditioned by voicing of the initials, splits conditioned by aspiration, glottalization, and frication have also been reported for Tai languages (Gedney 1989b; L-Thongkum 1997). However, in reconstructing initial consonants, knowing what the original voicing specification of those initials was is generally sufficient. Therefore, this dissertation adopts Li (1977)’s convention, in which 1 is the assigned code for tones developed from original voiceless stops and 2 for those developed from voiced stops. When there is a need to allow for more fine-grained categorization of the proto-initials, additional notation will be introduced. Note that the notation described above only applies to forms in Tai dialects. Reconstructed forms from languages other than Tai are presented in their original notation. Also, note the use of Chinese characters in this dissertation. While simplified characters are used generally in this dissertation, traditional characters are used in discussions regarding etymology of those characters. For example, the Tai dialect of Lungchow is written in Chinese with simplified characters 龙州 Lóngzhōu but the Chinese etyma for ‘needle’ and ‘pair’ are written with traditional characters 鍼 zhēn and 雙 shuāng respectively. 25 1.5 Overview of dissertation The remainder of this dissertation consists of six chapters. The next chapter, Chapter 2, addresses the issue of the PT prosodic word while the following three, Chapters 3, 4, and 5, deal with the segmental components of the PT prosodic word: the onset and the rime. These are followed by a chapter that addresses the issue of the PT tones, Chapter 5. The last chapter then discusses the contributions and implications of this dissertation. An overview of each chapter is as follows: Chapter 2 discusses the structure of the PT prosodic word. I first discuss the importance of the prosodic word in the reconstruction of PT. Then, I show that the monosyllabic view espoused by the conventional reconstruction fails to account for the range of comparative data and argue that PT was in fact a sesquisyllabic language. I also provide a brief typological survey of the canonical shape of the prosodic word in SEA languages. Lastly, I claim that PT was a sesquisyllabic language whose prosodic word can be either monosyllable or sesquisyllable and then outline the phonotactics of PT onsets, rimes, and tones. Chapter 3 deals with the PT onset system focusing on simple onsets. I propose a system of PT simple onsets that departs from the conventional reconstruction in a number of major respects. First, I reconstruct a contrastive series of uvular consonants in addition to the five places of articulation in Li’s system. Second, I argue that the so-called “voicing alternation” is in fact two distinct phenomena—borrowings and reduction of sesquisyllables. Last but not least, I claim that PT did not have contrastive aspiration. Chapter 4 takes up the issue of complex onsets in PT. I first show that the monosyllabic hypothesis with only canonical clusters fails to account for the existing range of onset correspondences. I then present the inventory of reconstructible PT 26 complex onsets that includes two types of PT complex onsets: tautosyllabic clusters and sesquisyllabic clusters. Lastly, I show that the process through which Tai languages became monosyllabic cannot be viewed as a unified process of monosyllabification. Chapter 5 discusses the rime system of PT. I first argue that PT had a fullfledged vowel length contrast which interacted closely with syllable structure. I account for the so-called “Gedney’s puzzle” by positing four sets of rimes that do not add any complexity to the system. Lastly, I propose that final *-l, *-c, and perhaps *-ɲ also existed in PT in addition to those codas assumed in the conventional reconstruction. Crucially, the proposed system takes seriously the effect of medial and coda consonants on the development of the vowels. Chapter 6 proposes a concrete reconstruction of the PT tonal categories *A, *B, *C, and *D, already well-established in the conventional reconstruction. I argue that the contrast between PT tones was based not solely on pitch but also on voice quality. In particular, I posit a modal tone with a mid level contour for *A, a creaky tone with a low rising contour and a relative long vowel duration for *B, and a high falling contour ending with a glottal constriction with a relatively short vowel duration for *C. As for checked syllables, I posit a low rising contour for *D. Finally, Chapter 7 discusses the contributions of this reconstruction. I first compare the proposed reconstruction of PT phonology with Li’s conventional reconstruction. Then, I propose a preliminary subgrouping schema of Tai languages as well as a historical scenario that explains the diversity and the convergence among modern Tai dialects. 27 CHAPTER 2 THE PROTO-TAI PROSODIC WORD 2.1 Introduction Never seriously discussed in historical studies of Tai languages is the structure of the PT prosodic word. In reconstructing Proto-Tai, the focus has always been on reconstruction of individual segments to account for individual correspondences. This lack of attention to the organization of segments into larger units has ironically hindered progress in solving the mysteries in segmental reconstruction. This is because in all natural languages individual segments must be organized within larger domains such as syllables, and words. These prosodic domains crucially pose certain constraints on how segments are combined. Therefore, reconstruction of segments and consequently individual etyma depends largely on the researchers’ implicit or explicit assumptions about possible phonological shapes of prosodic words in the protolanguage. A well-known example is the case of the *DVD constraint on the Proto-IndoEuropean (PIE) root, which has been used as important evidence for the Glottalic Theory (Gamkrelidze & Ivanov 1972). In the Standard Theory of PIE stops (Mayrhofer 1986), PIE phonology had three contrastive phonation types in stop consonants—voiceless, plain voiced, aspirated voiced. These stops, when combined, yield nine logically possible CVC root types. Nevertheless, it has been generally assumed that there was a constraint against CVC roots having two plain voiced stops, i.e. *deg (Iverson & Salmons 1992). The Glottalic Theory, in contrast, believes that the three contrastive series of stops in PIE was in fact (aspirated) voiceless, ejectives, and (aspirated) voiced. In support of this segmental reconstruction, the Glottalic Theory claims that the *DVD constraint is common in ejective languages. In this view 28 *DVD root would have in fact been *T’VT’, which is not allowed in ejective languages. This putative *DVD constraint has however been challenged by various Indo-Europeanists such as Iverson and Salmons (1992), Barrack (2002), and Cooper (to appear). As the root is an obligatory part of a prosodic word, this example from Indo-European illustrates how assumptions about possible phonological shapes of the prosodic word, or some part of it, could influence the reconstruction of the segments. Another example closer to Tai is the case of Old Chinese (OC). Karlgren (1964), Li (1980) and Baxter (1992) all assume that the OC was a monosyllabic language, while Sagart (1999), and Handel (1998) argue that OC allowed for sesquisyllabic words. The different hypotheses about the shape of the OC prosodic word have led to major differences in segmental reconstruction. An example comes from cases of etyma whose Middle Chinese (MC) onsets point to OC *t- or *th-, but whose word-family connections point to another initial consonant. Assuming that OC was strictly monosyllabic, Karlgren (1964) reconstructs *təp and *t̑ ’i̯ wət for 答 dá ‘to respond’ and 出 chū ‘to exit’ respectively. In contrast, Sagart (1999: 90), who holds that sesquisyllabic words were common in OC, posits *at-gɨp and *bt-khut for the two etyma. This OC case is especially relevant not only because of the geographical proximity to Tai but also because a case can also be made for reconstructing PT either as a monosyllabic or sesquisyllabic language. Therefore, it is crucial that possible shapes of the PT prosodic word be established before proceeding to reconstruct the onset, the rimes, and the tones. In this chapter, I first review the structure of the prosodic word as assumed by earlier proposals on PT reconstruction and show that the conventional view that PT words were all monosyllables cannot account for the comparative data. Crucially, I argue that the view that PT was a sesquisyllabic language is the best hypothesis regarding the PT prosodic word. Then, I briefly discuss the prosodic word and its typology from the 29 metrical and prosodic points of view (Hayes 1995; Liberman 1975; McCarthy & Prince 1986; Nespor & Vogel 1986; Ussishkin 2005). Lastly, I put forward a proposal establishing an explicit characterization of the PT prosodic word. In particular, I claim that PT prosodic words can be either monosyllables, or sesquisyllables. Each word contained exactly one iambic foot headed by a heavy syllable. The phonation-type and place contrasts in the onset were neutralized in the coda position. Lastly, each prosodic word is specified for one lexical tone. 2.2 The prosodic word In the framework of Prosodic Phonology (Inkelas 1989; McCarthy & Prince 1986; Nespor & Vogel 1986; Selkirk 1981), the prosodic word is described as “the lowest constituent of the prosodic hierarchy which is constructed on the basis of mapping rules that make substantial use of non-phonological notions (Nespor & Vogel 1986: 109).” It is also characterized formally as the prosodic category that is larger than the foot but smaller than the phonological phrase. It is the phonological domain that roughly corresponds to the morphological word, and thus the smallest prosodic constituent that can stand alone (Dixon & Aikhenvald 2002: 24-25; Zec 2005)5. Establishing the PT prosodic word, therefore, means figuring out what possible PT words may look like phonologically. In this section, I first discuss the importance of the prosodic word in synchronic phonology. Then, I outline some well-formedness conditions relevant to the discussion of the PT prosodic word. 5 According to Dixon and Aikhenvald (2002: 24-25), a morphological word can stand alone only when it is also a prosodic word. For examples, lexical words and free function words in standard Serbian can stand alone and carry a pitch accent but bound function words must be cliticized to a host prosodic word (Zec 2005). 30 2.2.1 The importance of the prosodic word The prosodic word is very important from a synchronic perspective. It is the domain in which rules and/or constraints are applied to maintain the syllables and the feet in conformity with universal or language-specific phonological principles. Finnish, for instance, requires that vowels within a prosodic word be all front or all back (van der Hulst & van de Weijer 1995: 498). Examples of permissible words are väkkärä ‘pinwheel’ and pöytä ‘table’, both of which have only front vowels. Similary, makkara ‘sausage’ and pouta ‘fine weather’ exemplify Finnish words with back vowels. This means that words like *väkkara or *poutä do not exist in the language. Some other examples of rules that make references to the prosodic word include vowel harmony in Hungarian and stress assignment in Latin (Nespor & Vogel 1986). Furthermore, the prosodic word is important in synchronic phonology because individual languages impose certain phonotactic requirements on their prosodic words. The Austroasiatic language Kammu described by Svantesson (1983) is a good example. In this language, prosodic words must contain exactly one iambic foot (Pittayaporn 2005c). In other words, a simple Kammu word must either be a heavy monosyllable, i.e. /jìm/ ‘red’ and /cè:m/ ‘wet’, or a sesquisyllable, i.e. /kʔí:p/ ‘centipede’ and /k.màʔ/ ‘rain’. Words like */bi/ or */tu/ are not permissible because they only have one short vowel, which makes them too light for a Kammu prosodic word. Similarly, non-compound words like */(pə)(təban)/, or */(sa:)(kətum)/ do not exist in Kammu because they consist of two feet6, exceeding the weight maximum for the Kammu prosodic word. In addition to the case of Kammu, some additional examples of phonotactic requirements imposed on the prosodic word include the ban of lax vowels at the end the prosodic word in German (Hall 1997) and the constraint in 6 The foot is represented here by ( ). 31 Dutch that within a prosodic word the first consonant in a heterosyllabic consonant sesquence has to be more sonorous than the second (Booij 1997). In short, the prosodic word is a prosodic domain that plays an important role in determining word-level phonotactics and other phonological phenomena within the word. Any comparative reconstruction therefore needs to take seriously the shape of the prosodic word in the proto-language. For example, if we were to reconstruct Finnish, we must therefore not posit words that contain both front and back vowels. Similarly, if we were to reconstruct Kammu, we must include only heavy monosyllabic words and sesquisyllabic words in the reconstructed lexicon. 2.2.2 Well-formedness A well-formed prosodic word in a given language is one that simultaneously satisfies all the well-formedness requirements of that language including the wellformedness requirements imposed on the syllables (Goldsmith 1995). In this subsection, I discuss three most relevant phonological concepts: size requirements, sonority, and foot typology. 2.2.2.1 Size requirements Numerous languages of the world exhibit restrictions on the size of possible prosodic words. Many impose restrictions on the minimal size a prosodic word can have, while fewer prohibit prosodic words that exceed a certain size. These types of restrictions can be referred to as word minimality and word maximality, respectively. These size requirements should not be thought of as primitive phonological conditions but should be characterized as effects of interactions among various prosodic and metrical constraints (de Lacy 2004; McCarthy & Prince 1986). 32 Between the two types of size requirements, word minimality is found more generally cross-linguistically. Although cross-linguistic surveys have shown that there is no strong correlation between the minimal word size and the minimal foot (Downing 2006:48-58, 94-100), diverse languages from all over the world require that their prosodic words be minimally one well-formed metrical foot. In SEA, a number of languages including Burmese (Green 2005), Thai (Bennett 1994, 1995), Kayah Li (Bennett 1995), Khmer (Huffman 1972), Indonesian (Cohn & McCarthy 1998) and Moken (Pittayaporn 2005a) prosodic word have been shown to contain at least one well-formed bimoraic foot. The maximality requirement is not found as commonly as its minimality counterpart. Maximality effects have been shown to operate in Maori and perhaps in Ura (de Lacy 2004). Although unfortunately very few studies have been done on maximality, especially in Southeast Asian languages, some languages including Burmese (Green 2005), Kammu (Pittayaporn 2005c), Moken (Pittayaporn 2005a) and Indonesian (Cohn in progress) have been shown to have at least a strong dispreference for prosodic words that exceed one metrical foot. Although literature on interaction between word minimality and word maximality is lacking, many languages seem to show effects of both resulting in a situation where the well-formed word is equal to a well-formed foot. For example, in Maori, word minimality requires that a prosodic word be at least one metrical foot while word maximality bans any prosodic words that are larger than one foot. As a result of the interaction, normal prosodic words in Maori contain one single foot and optionally one unfooted syllable. Prosodic words with two feet are allowed only under special circumstances (de Lacy 2004). In Modern Hebrew, prosodic words optimally consist of two syllables that form one metrical foot. Trisyllabic words are allowed only to realize trisyllabic affixes (Ussishkin 2005). Similarly, in Khoekhoe, lexical 33 words must be exactly one metrical foot (Brugman 2009) due to the joint effect of minimality and maximality. In SEA, languages that require their prosodic word to be exactly one metrical foot are very common. Some clear cases of such languages include Kammu (Pittayaporn 2005a), Moken (Pittayaporn 2005a), and, to a large extent, Burmese (Green 2005) and Indonesian (Cohn in progress). 2.2.2.2 Sonority A well-formed prosodic word is one that can be divided exhaustively into one or more well-formed syllables (Booij 1997). One of the ingredients in the wellformedness of the syllables and the prosodic word is sonority. In particular, strings of segments are syllabified into syllables according to constraints on the relative sonority of the segments. These syllables are then combined to form the prosodic word. Although its definition is an issue of much debate, it is clear that sonority is linked at least partly to intensity or loudness (Clements 1990; Parker 2002). Sonority can be understood, and can be characterized in terms of hierarchy or scales. Many versions of this hierarchy have been proposed (e.g. Hooper 1976; Selkirk 1984; Steriade 1982), but in this dissertation I adopt Clements (1990)’s proposal that arranges non-syllabic segments from less sonorous to more sonorous classes as given Figure 2-1. Obstruent 0 Nasal 1 Liquid 2 Glide 3 Figure 2-1 Sonority scale (based on Clements 1990: 292) In the scale given here, obstruents are the least sonorous and can be said to have the sonority value of 0. On the opposite end of the scale, glides are the most 34 sonorous among non-syllabic segments and have the sonority value of 3. Nasal and liquid consonants occupy intermediate positions along the scale, having the sonority values of 1 and 2 respectively. These sonority values are very useful in describing the sonority profile of segments or string of segments. Most relevant to the issue at hand is the relative sonority among consonants that form a complex onset in a well-formed syllable. Cross-linguistically, permissible consonant clusters are subject to an important constraint, generally known as the Sonority Sequencing Principle (SSP), which states that “in any syllable, there is a segment constituting a sonority peak that is preceded and/or followed by a sequence of segments with progressively decreasing sonority values (Selkirk 1984: 116).” In other words, in a given syllable sonority is lowest at the edges and highest at the syllabic nucleus. Because well-formed syllables are subject to SSP, as part of the syllable complex, onsets are generally expected to follow the same principle as well. The role of SSP in well-formedness of complex onsets can be illustrated by the case of Spanish initial clusters. Spanish strictly obeys SSP so that all permissible complex onsets consist of a stop followed by a liquid or a glide. For instance, /kl-/, /kr-/ and /kw-/ are well-formed initial clusters as in /klabe/ ‘key’, /krisis/ ‘crisis’, and /kwota/ ‘quota’. This is because /k-/ has lower sonority than /-l-/, /-r-/ and /-w-/, which are closest to the syllabic nucleus. In contrast, /lk-/, /rk-/, /wk-/, are not allowed because the /k-/ is less sonorous than /-l-/, /-r-/, and /-w-/ violating SSP (Harris 1983; Hualde 1991). In addition to SSP, languages may further impose a minimal sonority distance on complex onsets (Levin 1985; Selkirk 1984; Venneman 1972; Zec 2007). Such constraint requires that members of complex onsets meet a certain difference in sonority values. Recall the sonority values of different types of consonants in Figure 2-1. In Spanish, /kl-/, /kr-/ are well-formed clusters because the sonority distance between the 35 obstruent /k-/ and the following liquids is exactly 2. Similarly, the difference in sonority values between the initial obstruent /k-/ and the glide /w-/ is 3, exceeding the required minimal sonority distance. On the other hand, */pn-/ and */ml-/ are banned in Spanish, even though they do not violate SSP. This is because the sonority distance within each of the consonant sequences is only at 1 respectively (Harris 1983). Although the role of sonority in syllable well-formedness is robustly attested across languages, clusters that are not predicted by sonority-based generalizations are relatively common. The most well-known case is English clusters /sp-/, /st-/, and /sk-/ as in spy, star, and sky respectively. In these clusters, the /s-/ and the following stops form sonority plateaus in violation of SSP, but they are still well-formed (Halle & Vergnaud 1980; Steriade 1982). Modern Hebrew allows biconsonantal clusters that consist of an obstruent followed by another obstruent, e.g. /dg-/ as in /dgalim/ ‘flag’, /ks-/ as in /ksamim/ ‘magic’, and /pg-/ as in /pgarim/ ‘carrion’. Instead of a rising sonority, such clusters form sonority plateaus as both the first and the second consonants have the sonority value of 0. A more extreme case is Polish, which allows long strings of segments that violate SSP to precede the vocalic nucleus, e.g. bdzura ‘nonsense’, fstręt ‘repulsion’. Interestingly, these complex onsets in Polish have been analyzed as consisting of two separate onsets (Gussmann 1992) or a degenerate syllable followed by an onset (Cho & King 2003). Sesquisyllabic structure (see §2.4.3) found widely in SEA has been similarly analyzed as comprising a degenerate syllable followed by an onset (Cho & King 2003; Pittayaporn 2005c). Because in these cases, the consonants in each sequence are not all linked to the same syllable nodes, they in fact do not form syllable clusters in the strict sense. Kreitman (2008) surveys 62 languages with complex onsets and concludes that there is an implicational hierarchy for the occurrence of different types of consonant 36 clusters (S = sonorant, O = obstruent): SO  SS  OO  OS. According to this typology, OS clusters, e.g. /pr-/, /kl-/, /kw-/ etc. are the most basic type of clusters. The presence of OO clusters, e.g. /pg-/, /st-/, /kt-/ etc. implies the presence of SO clusters. Similarly, the presence of SS clusters, e.g. /ml-/, /nl-/, /rl-/ etc., implies the presence of OS and OO clusters. Lastly, the presence of SO clusters, e.g. /rt-/, /ld-/ etc., implies the presence of the other three cluster types. In this dissertation, I refer to OS clusters as “canonical” or “unmarked” clusters, and refer to the other types of clusters as “marked” clusters. 2.2.2.3 Foot typology In Prosodic Phonology (McCarthy & Prince 1986; Nespor & Vogel 1986), a prosodic word is made up of one or more metrical feet, each of which is made of one or more syllables. By transitivity, a well-formed prosodic word consists of one or more well-formed syllables. According to Hayes (1995), cross-linguistically a wellformed foot belongs to one of the following three types: syllabic trochee, moraic trochee, and iamb. This foot typology is based on quantity sensitivity and dominance. The three basic foot types can be illustrated in Figure 2-2. 37 Left dominant Quantity-insensitive syllabic trochee ( .) σσ Quantity-sensitive moraic trochee (̮ ̮.) or (ˍ) Right dominant iamb ̮ ˍ(. ) or () σ  = stressed syllable . = unstressed syllable ̮ ˍ= light syllable = heavy syllable Figure 2-2 Foot typology according to Hayes (1995) A syllabic trochee is a metrical foot with initial prominence that counts syllables regardless of their weight (Hayes 1995). Pintupi is an example of languages that have this type of foot as illustrated by the following words: /('paɳa)/ ‘earth’, /('maɭa)('wana)/ ‘through from behind’. The former word consists of one syllabic trochee while the latter consists of two (Hayes 1995). Among Southeast Asian languages, Indonesian is an example of a language that has syllabic trochees (Cohn & McCarthy 1998). A moraic trochee refers to a metrical foot that consists of two moras and, similar to a syllabic trochee, has initial prominence. A moraic trochee can either consist of two light syllables or one heavy syllable (Hayes 1995). Cairene Arabic is a good example of a language that have this type of foot as illustrated by the following words: /(kata)('bitu)/ ‘she wrote it (m.)’ and /(ʔad)(wiya)('tuhu)/ ‘his drugs (nom.)’. In the former each of the two feet had two light syllables while the latter consists of one 38 foot with one single heavy syllable followed by two feet of two light syllables (Hayes 1995). In SEA, Tondano, spoken in Sulawesi, is an example of languages with moraic trochees (Ball 2003). Lastly, an iamb is a metrical foot with final prominence. It can either consist of two syllables with a light first syllable, or one single heavy syllable (Ball 2003). In Creek, the prosodic word consists of iambic feet as exemplified by the following words: /(ni'ha:)/ ‘lard’ and /(ti:)(ni:t)(ki:)/ ‘thunder’. While the first word consists of one iambic foot with a light syllable followed by a heavy one, the second word consists of three feet, each of which is a heavy syllable (Hayes 1995). In SEA, numerous languages belong to this type, including Thai (Bennett 1994, 1995), Kayah Li (Bennett 1995), Burmese (Green 2005), Moken (Pittayaporn 2005a), Kammu (Pittayaporn 2005c; Svantesson 1983), and others. 2.3 Reconstructing the PT prosodic word Assumptions about the canonical shape of the PT prosodic word are a central factor underlying the controversies about the reconstruction of PT. Unfortunately, they are not usually discussed and to my knowledge not explicitly compared. Although the conventional view represented by Li (1977) among others holds that PT was monosyllabic, alternative proposals have also treated PT as sesquisyllabic (Ferlus 1990), or disyllabic (Haudricourt 1975; Strecker 1983). Advances in the study of Kra-Dai languages have shown that Proto-Kra-Dai cannot have been a strictly monosyllabic language. Evidence for such a conclusion comes from comparative studies of lesser-known relatives of Tai. Ostapirat (2000) reconstructed Proto-Kra with sesquisyllables on the basis of onset correspondence and tonal reflexes in modern languages. For example, he posits *p.laA ‘afraid’ to account for Gelao /lauA2/ and Paha /pja:A2/, in contrast to *platD ‘blood’ for Gelao /plɒD1/ and 39 Paha /pɛ:D1/. While reflexes of the sesquisyllabic *p.l- have tone in the second series according to the voicing of the non-initial *.l-, the reflexes for the tautosyllabic cluster have tone in the first series according to the initial *p-. Edmondson and Yang (1988) present initial correspondences among Kam-Sui languages as evidence for “preconsonants” which perhaps reflect disyllabic or sesquisyllabic structure. For example, Mulam /khɣətD1/ corresponding to Kam /sətD1/ suggests Proto-Kam-Sui *kjətD ‘tail’. Similarly, Ferlus (1989; 1996b) shows convincingly that Proto-Kam-Sui was a sesquisyllabic language although all modern Kam-Sui languages are monosyllabic. For example, he reconstructs *k.s- as onset of Kam-Sui etymon ‘sour’ which is now reflected as /səmC/ in Kam, /humC/ in Sui, and /khɣəmC1/ in Mulam. Similarly, L-Thongkum (1992) proposes a scenario in which Lakkja nasalized vowels developed from earlier sesquisyllabic forms, e.g. /kʰũA1/ ‘pig’ from earlier *kh-mu. Further, Osaka (2007) proposes that Be /vəiA2/, Siamese /pi:A1/ and Saek /pi:A1/ comes from Proto-Be-Tai *Cə̆ pii ‘year’. This growing consensus that prosodic words in Proto-Kra-Dai were not all monosyllabic is clearly at odds with the strict monosyllabicity of cognates found among modern Tai languages. This raises the issue of what the situation at the PT stage was. In other words, it is still unclear whether the Proto-Kra-Dai prosodic word had reduced to monosyllables by the PT period. In this section, I show that the monosyllabic view of PT fails to account for the range of comparative data and argue in favor of the sesquisyllabic hypothesis. 2.3.1 The PT prosodic word in earlier proposals The conventional view of the PT sound system holds that the proto-language was strictly monosyllabic. Li does not explicitly state in the Handbook of Comparative Tai (Li 1977) that PT was strictly monosyllabic, but all of his reconstructed forms are 40 monosyllables. Other linguists working on PT reconstruction also share this view despite their disagreement about segmental reconstruction. Nishida (1954; Nishida 1955), Liang and Zhang (1996), Luo (1997), and Sarawit (1973) all posit only monosyllables in their reconstruction. Gedney, who rarely gives precise reconstructions of PT “morphemes,” always talk about PT in terms of syllables (Gedney 1972a, 1989b). Although the motivation for the monosyllabic view has never been discussed, it is safe to say that it stems from the fact that cognates found in modern Tai varieties are almost always monosyllables (Gedney 1989b: 15). This view clearly affects segmental reconstructions proposed by different authors. For example, Li (1977) proposes PT *dl- to explains cases in which some daughter languages indicate PT *lbut others indicate PT *r-. Li’s choice is motivated by the fact that *dl- is the only canonical alveolar cluster that he has not reconstructed. Similarly, Liang and Zhang (1996) reconstruct a marked cluster *xkr- to account for etyma that Li considers to be a special development of his *kʰl-. An alternative to the monosyllabic view is to reconstruct sesquisyllabic words for PT. The term “sesquisyllable”, literally “one-and-a-half syllable”, was originally coined by Matisoff (1973) to refer to words that consist of an unstressed “minor” syllable followed by a fully stressed “major” syllable. Crucially, the minor syllable is characterized as lacking a contrastive vowel, e.g. Palaung (Shorto 1960) /k.tɛ/ ‘earth,’ and /kr̩ .taʔ/ ‘tongue’7 (see §2.4.3). Many languages in SEA have been described as showing sesquisyllabic words, e.g. Khmer (Henderson 1952), Kammu (Svantesson 1983), Palaung (Shorto 1960), Burmese (Green 2005), Lahu (Matisoff 2001) etc., but the precise formal characterization of sesquisyllables is unclear. In fact, it is likely that 7 Shorto transcribe these forms as /kətɛ/ and /kərtaʔ/ respectively but states that “a minor syllable has no vowel other than an anaptyctic one.” 41 the so-called sesquisyllables differ structurally from one language to another. In this dissertation, I define a sesquisyllable as a prosodic word that consist of two syllables, the first of which is unstressed and lacks a phonological vowel. Under this definition, the sesquisyllabic structure differs from tautosyllabic clusters in that the consonants in a sequence at the beginning of a sesquisyllable are not all linked to a single syllable node. Ferlus (1990) proposes a reconstruction of the PT system of initial consonants that includes both monosyllabic and sesquisyllabic etyma8. This view leads him to interpret many correspondence sets differently from Li. For example, he posits *k.t- for etyma for which Li proposes *tl-. The reflexes in daughter languages, however, can be accounted for equally well by the two accounts. In addition, the sesquisyllabic view also faces the same challenge posed by the reconstruction of vowels. Since the minor syllable of a sesquisyllabic word does not have a contrastive vowel, the problematic vowel correspondences must be accounted for in the rime of the major syllables only. Another alternative holds that PT was disyllabic. Haudricourt (1956; 1963b) proposes that disyllabic roots be reconstructed for PT to account for some onset correspondence sets (see Chapter 4). For example, he links the Tai etyma for ‘bird’ and ‘water’ to Malay /danum/ and /manuk/ to explain the fact that the medial /-n-/ is reflected in these two etyma as /n-/ in Siamese but /r-/ in Yay, a NT—/na:mC1/ and /ramC1/ respectively. Furthermore, Haudricourt (1975) proposes that disyllabic roots be reconstructed for PT to account for the so-called “Gedney’s Puzzle (see Chapter 5).” For instance, he proposes that PT *apui ‘fire’ went through two different paths in different subgroups: 1) *apui > *apwi > *pwai > *vai in SWT and CT but 2) *apui > *apwi > *pwi > *vi in NT. However, it is not clear whether according to this account monosyllabic words were allowed in PT. Similar to the sesquisyllabic view, this disyllabic view seems to be motivated by the need to account for segmental correspondences, rather than by 8 He refers to sesquisyllables as “disyllables”. 42 typological considerations as is the case for the monosyllabic view. Examples of reconstructed forms under different assumptions are given in Table 2-1 for comparison. Table 2-1: Reconstruction of certain etyma under different assumptions about the PT prosodic word ‘fire’ ‘bear’ ‘moon’ ‘bird’ Li *vɛjA *ʰmwiA *ˀblɯənA *nrokD Ferlus9 *vɛjA *ʰmwiA *ɓlɯənA *C.nokD Haudricourt10 *apwiA *tumajA *bulanA *manukD In this section, I have reviewed the three views regarding the canonical shape of the PT prosodic word. All of them basically constrain how the segmental materials are to be reconstructed. In the following section, I will argue that the sesquisyllabic view is the best hypothesis regarding the PT prosodic word. 2.3.2 Sesquisyllabicity in PT The monosyllabic view of PT that underlies the conventional reconstruction has contributed enormously to the understanding of the history of Tai languages. However, it cannot account for the whole range of comparative data, and therefore fails to provide an accurate picture of PT phonology and its development into daughter languages. In particular, many onset correspondences show traces of PT words that 9 Ferlus (1990) does not reconstruct PT rimes. The rimes given in this table are taken from Li (1977). 10 Except for ‘fire’, Haudricourt does not propose PT forms. Forms found in this table are provided for illustration only. 43 were larger than monosyllables. Specifically, there is too wide a range of onset correspondences to be accounted for by positing a monosyllabic template for the PT prosodic word. Although these data are problematic for the monosyllabic hypothesis, they can be captured quite straightforwardly by the sesquisyllabic view. The first set of onset correspondences that is problematic for the monosyllabic view of the PT prosodic word is that in which most dialects point to PT *ɗ- (see §3.6.2.2), except for Saek. In Table 2-2, Siamese represents SWT, Lungchow represents CT, and Yay together with Saek represent NT. Li’s PT and the reconstructed Proto-Southwestern Tai (Pittayaporn 2008b, to appear-b) onsets are also provided for comparison. This correspondence cannot be accounted for if the reconstruction of PT is limited to monosyllables with unmarked clusters. Table 2-2: Correspondences involving PT *ɗ- ‘handle’ 1 ‘to acquire’ ‘single, only one’ ‘red’ 2 ‘bone’ ‘raw’ ‘star’ 3 ‘to transplant’ ‘earthworm’ PSWT *ˀd*ˀd*ˀd- Siamese da:mC1 da:jC1 diəwA1 dɛ:ŋA1 du:kDL1 dipDS1 da:wA1 damA1 dɯənA1 Lungchow da:mC1 dajC1 de:wA1 dɛ:ŋA1 dukDL1 dipDS1 da:wA1 damA1 dɤ:nA1 Yay da:mC1 dajC1 dewA1 diŋA1 dokDL1 dipDS1 da:wA1 damA1 dɯənA1 Saek da:mC1 dajC1 diəwA1 ri:ŋA1 rɔ:kDL1 ripDS1 tra:wA1 tramA1 truəlA1 Li *ˀd- *ˀdr- Li (1973; 1977: 107-110, 129-131) reconstructs PT *ˀd- for the first set and *ˀdr- for the second and third sets. This set of data has also been observed by Gedney 44 (1989c). With respect to the contrast between the second and third sets, Li (1977: 129) simply states that *ˀdr- gives /r-/ in Saek but is occasionally reflected as /tr-/. If we assume that the PT prosodic word was strictly monosyllabic with only canonical clusters, we have no way of explaining why Saek shows the contrast and why the stop in the third set is a plain voiceless instead of a glottalized stop. On the other hand, the sesquisyllabic view accounts for this paradigm by positing a coalescence of sesquisyllabic structure *t.n- into *ˀd- for set 311. The second set of correspondences that cannot be accounted for by the monosyllabic hypothesis with unmarked clusters, shown in Table 2-3, involves the uvular consonants (see §3.3). In addition to the NT language Saek, Siamese, Lungchow, and Yay data represent SWT, CT, and NT respectively. Li’s PT and PSWT reconstructions are given for comparison. Table 2-3: Correspondences with *q- ‘to kill’ 1 ‘torn’ ‘arm’ ‘to hang’ 2 ‘face down’ ‘soul’ ‘leg’ 3 ‘to sell’ PSWT *q- *qw*q- Siamese kʰa:C1 kʰa:tDL1 kʰɛ:nA1 kʰwɛ:nA1 kʰwamC1 kʰwanA1 kʰa:A1 kʰa:jA1 Lungchow kʰa:C1 kʰa:tDL1 kʰe:nA1 kʰumC1 kʰwanA1 kʰa:A1 kʰa:A1 Yay ka:C1 ka:tDL1 cenA1 venA1 hamC1 vanA1 ka:A1 ka:jA1 Saek ka:C1 ka:tDL1 ke:nA1 ve:nA1 (ŋamC1) hɔnA1 kwa:A1 kwa:jA1 Li *kʰ- *xw*kʰ- 11 Dempwolff and Dyen reconstructs this form as PAN *taNem ‘to plant’ (Wurm & Wilson 1975: 153) but it does not appear in Formosan languages. 45 The above table shows three correspondences that must be reconstructed with PT *q- (see §3.6.1.5)12. The first set is quite straightforwardly accounted for by positing PT simple *q-. The uvular stop is reflected as /kʰ-/ in most SWT and CT dialects including Siamese and Lungchow, and as /k-/ in NT. The second set should go back to PT *qwalthough the NT reflexes at first glance seem irregular. The usual reflex of this uvular cluster is /kʰw-/ in CT and SWT dialects and /v-/ in NT. Yay /hamC1/ ‘face down’ lost the medial /-w-/ early probably due to the labial quality of the final /-m/. The sequence *-wain ‘spirit’ thus must have become /ɔ/ in Saek before *ʰw- became *w-, i.e. PT *qwanA > *ʰwanA > hɔnA. The third set of correspondences shows unexpected reflexes in Saek, as noted by Haudricourt (1952), Li (1965; 1977: 195), and Gedney (1989c). Their reflexes in the rest of Tai point unequivocally to *q- but Saek shows a labial element in its reflex. It does not help to reconstruct any cluster with */-w-/ because medial */-w-/ are normally kept in daughter languages, cf. Siamese /kʰwanA1/ ‘spirit’. If PT prosodic words must be monosyllabic with unmarked clusters, these two forms must be treated as an unsystematic development as assumed by Li (1977: 195). However, data from the Kam-Sui branch of the Kra-Dai family indicate that the labiality of the Saek clusters is a retention. The modern Sui forms for ‘leg’ and ‘to sell’ are /pa:1/ and /pe1/ respectively (Institute of Language and Culture for Rural Development & Thailand and the Kam-Tai Institute 2003; Li 1965; Thurgood 1988). In contrast, if we assume that PT is sesquisyllabic, we are allowed to posit */p.q-/ for these two etyma. Note that Ostapirat (2005) connects PT ‘leg’ to PAN *paqa ‘thigh’. Note that the etyma in set 2 are not found in Kam-Sui. The third paradigm that argues against the monosyllabic view involves complex clusters whose reflexes in some daughter languages indicate PT *br- (see §4.2.1.6) but others indicates PT *ɟ- (see §3.6.3.3). The data are given in Table 2-4. 12 For arguments for reconstructing PSWT uvulars, see Pittayaporn (to appear-b). 46 Table 2-4: Correspondences involving *br- and *ɟ- ‘to separate’ 1 ‘to deceive’ ‘deficient ‘machete’ 2 ‘evening meal’ ‘tomorrow ‘to hate’ 3 ‘to taste’ ‘to soak’ PSWT *br*br*ɟ- Siamese pʰra:kDL2 pʰra:ŋA2 pʰrɔŋA2 pʰra:C2 pʰrawA2 pʰru:kDL2 cʰaŋA2 cʰimA2 cʰɛ:A2 Lungchow pja:kDL2 pja:ŋA2 pjoŋA2 pja:C2 pjawA2 pjukDL2 caŋA2 cimA2 Yay pja:kDL2 pjoŋA2 sa:C2 sawA2 sokDL2 saŋA2 simA2 se:A2 Saek pʰra:kDL2 pʰrɔ:ŋA2 tʰa:C2 tʰɔ:kDL2 saŋA2 se:A2 Li *br*vr*ɟ- The reconstruction of set 1 and set 3 is straightforward. Li (1977) posits *br(see §4.2.1.6) for set 1, and *ɟ- (see §3.6.3.3) for set 3. However, set 2 presents a serious problem to the simple monosyllabic view. While Siamese and Lungchow reflexes point to PT *br-, Yay and Saek reflexes all point to *ɟ-. To account for this problematic correspondence, Li tentatively reconstructs *vr- for these etyma, but it is not clear how *vr- could have developed into a palatal stop. This correspondence, however, is not a problem for the sesquisyllabic view, which allow for complex structure *ɟm̩ .r- in PT. Under this view, *ɟm̩ .r- simplified to *mr-, and became *br- in Siamese and Lungchow. The cluster took a different path in Yay and Saek, where it simplified to *ɟr, and became *ɟ-. The last set of sound correspondences involves labial clusters. Not only does it pose a challenge for the view that PT was a monosyllabic language that only allowed unmarked clusters, but it also shows that PT could not have been a monosyllabic 47 language that had marked tautosyllabic clusters, e.g. *pt-, *qs-, either. The data in this paradigm is given in Table 2-5. Table 2-5: Correspondences involving labial clusters ‘vegetable’ 1 ‘to walk’ ‘head hair’ ‘eye’ 2 ‘to die’ ‘grasshopper’ ‘to be exposed’ 3 ‘to burst’ to hunt PSWT *pʰ*t*t- Siamese pʰakDS1 pʰa:jC1 pʰomA1 ta:A1 ta:jA1 takDS1 ta:kDL1 tɛ:kDL1 Lungchow pʰjakDS1 pʰja:jC1 pʰjumA1 ha:A1 ha:jA1 pʰja:kDL1 pʰe:kDL1 pʰjawB1 Yay piəkDS1 pja:jC1 piəmA1 ta:A1 ta:jA1 takDS1 ta:kDL1 te:kDL1 tawB1 Saek pʰrakDS1 pʰramA1 pra:A1 pra:jA1 pra:kDL1 pre:kDL1 Li *pʰr*tr*pr- Assuming the view that PT only had monosyllabic prosodic words with canonical clusters, Li (1977) reconstructs the data in set 1, set 2, and set 3 as *pʰr-, *tr-, and *pr- respectively. It should be immediately clear that such a view cannot account for the correspondences in set 2 and set 3. Specifically, both of these sets must have gone back to consonant sequences that contain a labial consonant as suggested by the fact that Saek has /pr-/ for both. The labiality in the NT language Saek in set 2 cannot be viewed as an innovation because the labial stop is also attested in the CT dialect of Bao Yen, e.g. /pʰja:A1/ ‘eye’, as well as in the Kam-Sui language Lakkja, e.g. /pla:A1/ ‘eye’. The simple monosyllabic view clearly cannot account for the distinction between set 2 and set 3, and would have to reconstruct *pr- for both. Although one may argue for positing *fr- for set 2, such an ad hoc solution can be easily ruled out. 48 Crucially, it is hard to imagine how *fr-13 could have become *t- in SWT and CT. Moreover, it will become clear in §3.6.9.3 that *f- did not exist in PT. In contrast, the sesquisyllabic hypothesis straightforwardly accounts for this paradigm of data. As Li’s *pʰr- must in fact be reconstructed as *pr- (see discussion on lack of contrastive aspiration in PT in §3.5), Li’s *tr- and *pr- can be reconstructed with *p.t- and *p.r- respectively. According to this scenario, the etyma in sets 2 and 3 must go back to two sesquisyllabic sequences-- *p.t- and *p.r-. These two PT onsets merged in some languages but are kept distinct in others. In Siamese and Yay, *p.tand *p.r- merged into *p.t-, and later simplified to /t-/. Similarly, Saek merged *p.tand *p.r- but now reflects the result of merger as /pr-/. In contrast, Lungchow keeps *p.t- and *p.r- distinct, and now shows /h-/ and /pʰj-/, respectively. In short, the sesquisyllabic view can successfully accounts for both labial reflexes without losing the distinction between set 2 and set 3. Although the disyllabic view can account for this paradigm as well as the sesquisyllabic view, the monosyllabic view with marked clusters fails to capture the distinction between the three sets of data. As will become clear in §3.5, PT did not have contrastive aspiration, meaning that Li’s *pʰr- was in fact *pr-. The elaborate monosyllabic view can posit *pr- for set 1, and *pt- for set 2, but it must leave set 3 unaccounted for. Because by definition all members of a true cluster must be linked to the same syllable node, a distinction between *pr- and *p.r- is thus ruled out. Essentially, the data in set 1 must be reconstructed with PT tautosyllabic *prcontrasting with set 3 which goes back to PT sesquisyllabic *p.r- respectively. One may propose an ad hoc solution that perhaps the etyma in set 2 may go back to marked tautosyllabic cluster *fr-, *fpr-, or even *fpt-, but as pointed out earlier PT did not 13 Li (1977) does in fact reconstruct *fr- but he uses it to account for a different set of data. 49 have *f-. Therefore, it is now clear that the view that PT was strictly monosyllabic but allowed marked clusters is also not tenable. Not only does the empirical evidence provided so far show that PT cannot have been strictly monosyllabic, it also shows that the sesquisyllabic hypothesis can effectively account for the whole range of comparative data. Unfortunately, empirical evidence that would allow us to compare the sesquisyllabic view against the disyllabic view does not exist. However, the sesquisyllabic hypothesis is much more plausible if the chronology of the shift toward monosyllabicism is taken into account. More specifically, the sesquisyllabic hypothesis offers a more plausible scenario of how modern languages evolved from the PT system. In SEA, the typical path of evolution is from disyllabicism to sesquisyllabicism, and then to monosyllabicism (Ferlus 1996a; Matisoff 2001). This evolutionary scenario has been amply attested in the different language groups of SEA, e.g. Chamic (Haudricourt 1956; Lee 1966; Thurgood 1999), Tibeto-Burman (Matisoff 2001), and Vietic (Ferlus 1975). Assuming this evolutionary path, it goes without saying that the monosyllabic hypothesis is the most straightforward of the three. Because all attested daughter languages of PT are monosyllabic at least in the inherited portion of the lexicon, we can derive modern monosyllabicity directly from a monosyllabic proto-language. Well-known cases of non-monosyllabic Tai languages are Siamese and Lao, but it is clear that polysyllabic words in these two languages are either loanwords—mostly from Indic languages (Gedney 1965) and Khmer (Varasarin 1984)—, or results of reduction of earlier compounds (Kullavanijaya 1992b). This fact alone would suggest that the hypothesis that the PT prosodic word was strictly monosyllabic is the most likely one. However, I have shown earlier that the monosyllabic view is insufficient as it fails to capture the range of comparative data. Therefore, the sesquisyllabic and 50 disyllabic view became the only remaining candidates for the characterization of the PT prosodic word. Although the two hypotheses conflict with the typological fact that all attested modern languages are monosyllabic, they still are not implausible. There are numerous cases of typological mismatches between a proto-language and the attested modern languages. Old Chinese, for example, was uncontroversially a nontonal language (Baxter 1992; Haudricourt 1954a; Pulleyblank 1962; Sagart 1999; Schuessler 2007) but no modern Chinese dialect attests such a typological profile. Between the two remaining hypotheses, the sesquisyllabic hypothesis offers a more chronologically plausible scenario than the disyllabic hypothesis. The sesquisyllabic view has to assume that different branches underwent monosyllabicization separately. However, it is still not impossible because it only takes one evolutionary step to reduce sesquisyllables to monosyllables, and there are many well-documented cases. In contrast, the disyllabic view would have to assume that different subgroups went all the way from disyllabicity to monosyllabicity separately without any modern language that would attest intermediate, perhaps sesquisyllabic, steps. Therefore, the chronological plausibility renders the disyllabic hypothesis only marginally possible. In this section, I have shown that among the different views of the PT prosodic word, the sesquisyllabic view is the best hypothesis as it captures nicely the range of comparative data and at the same time offers a plausible scenario of the shift toward the monosyllabicism in modern dialects. In contrast, the monosyllabic view is at first glance the most plausible hypothesis, if we only take the typology of modern Tai dialects into account. However, it fails to capture the whole range of sound correspondences among the dialects. On the other hand, the disyllabic can capture the comparative data quite nicely but it is very unlikely from the point of view of 51 chronological plausibility. Therefore, I argue that PT should be characterized as a sesquisyllabic language. 2.4 The prosodic word in Southeast Asia As an integral part of the Southeast Asian linguistic area, Tai languages in all probability must have experienced some of the convergent currents that have swept through the area. It is therefore important in establishing the canonical shape of the PT prosodic word to examine prosodic types attested in languages of SEA. With respect to the prosodic word, Matisoff (2001) views syllable structure and word structure of Southeast Asian languages as constantly changing concomitantly with emergence and disappearance of tonal contrast. The Chamic branch of Austronesian is an excellent case that illustrates this constant restructuring of the prosodic word. Proto-Chamic evolved from a disyllabic language with stress mainly on the penult into the various types of Chamic languages found today (Lee 1966). Today, Chamic languages are either monosyllabic, sesquisyllabic, or disyllabic but always with ultimate stress (Lee 1966; Thurgood 1999). This difference among these closelyrelated languages highlights the fact that the canonical shape of the prosodic word is a property that is susceptible to change. In this section, I review the three most common types of languages in SEA— monosyllabic, disyllabic, and sesquisyllabic. These types are defined by the canonical shape of their prosodic words. They will be discussed in relation to concepts relating to well-formedness in §2.2.2. It is crucial to remember that taking other languages in the area as models for reconstruction is not the same as forcing alien characteristics into the proto-language being reconstructed. Although the proto-language we are reconstructing may be structurally very different from other languages in the linguistic 52 area, areal typology helps contextualize the reconstruction and provide models against which the plausibility of the reconstruction can be checked. 2.4.1 Monosyllabic languages Monosyllabic languages refer to linguistic varieties whose prosodic words are canonically monosyllabic words. Examples of monosyllabic languages in SEA include Dehong Tai (Luo 1997), Nhahuen (Ferlus 1998), Vietnamese (Ðoàn 1992), Muong (Tài 1982), White Hmong (Ratliff 1992), Tsat (Thurgood 2005), among others. In these languages, both word maximality and word minimality must work together to ensure that each prosodic word is exactly one metrical foot. Logically a metrical foot can consist of one or two syllables, but obviously in monosyllabic languages a prosodic word canonically consists of only one syllable. According to the foot typology discussed in §2.2.2.3, only a heavy syllable can be a well-formed foot by itself. This means that the prosodic word in monosyllabic languages must be heavy. This indicates that there is an additional prosodic restriction in these languages that forces each syllable to be heavy. Although monosyllabic languages share the word size requirements, they differ with respect to foot type, and permissible onsets and codas. With regards to foot typology, a monosyllabic language in theory can be either trochaic or iambic. However, evidence for analyzing a monosyllabic language as trochaic or iambic is usually not available. With respect to codas and onsets, they may differ in what consonant combinations are permissible as syllable onsets, and what kinds of codas are allowed. Monosyllabic languages in SEA may or may not allow complex onsets, but in the majority of languages only canonical clusters exist. In other words, the clusters in the overwhelming majority of monosyllabic languages in SEA respect SSP 53 and show the sonority distance of at least 2. To my knowledge, Khmer is the only language that potentially has marked clusters (Henderson 1952; Nacaskul 1978). Some monosyllabic languages allow only simple onsets while others allow a wide range of consonant clusters. Dehong Tai (Luo 1999), spoken in Yunnan, is an example of a monosyllabic language that does not allow complex onsets. In addition to the constraints enforcing monosyllabicity, this Tai language also has a very strict constraint against complex onsets. As initial clusters do not exist in the language, the prosodic word in Dehong Tai is canonically CV(C) or CV(:)C. Table 2-6 provides the inventory of onsets in Dehong Tai. Table 2-6 Permissible onsets in Dehong Tai (based on Luo 1999: xiii) plain aspirated affricate fricative nasal liquid glide labial ppʰ- fm- w- alveolar ttʰtssnl- palatal j- velar k- xŋ- glottal ʔ- h- Nhahuen (Ferlus 1998) on the other hand allows two consonants to occur in the onset position, e.g. /cr-/ in /criəw/ ‘cliff’, /gj-/ in /gje:t/ ‘to kill’, /ɲw-/ in /ɲwɤ/ ‘bat’. Therefore, the Nhaheun word canon is C(C)V(C). Table 2-7 presents the inventory of complex onsets in Nhaheun. 54 Table 2-7: Complex onsets in Nhaheun (edited from Ferlus 1998: ii) voiceless stops -r- -l- -w- voiced stops -r- -l- -w- -j- aspirated stops -j- fricatives -r- nasals -r- -w- geminate sonorants fricatives -r- labial prpl- brbl- pʰj- mr- mmwr- alveolar tr- drdw- srnwnnll- palatal cr- ɟrɟw- ɲrɲwɲɲ- velar krklkwgrglgwgjkʰj- ŋrŋwŋŋ- The inventory in Table 2-7 shows that this dialect of Nhaheun, except for geminates, obeys SSP outlined in §2.2.2.2. This means that in this language the first consonant in a complex onset is expected to be less sonorous than the consonants following it. For example, /gr-/ is possible because /g/ is less sonorous than /r/, cf. /grah/ ‘to bite’. On the same principle, the tautosyllabic cluster /rg/ does not occur in this language because /r/ is more sonorous than /g/. Similarly, /dw-/ is a permissible complex onset because /d/ is less sonorous than /w/. The geminate nasals are apparent exceptions to this generalization. Unlike the onsets, the set of permissible codas in each language seems unrelated to the fact that the language is monosyllabic. While Vietnamese only allows 55 final stops, nasals, and glides (Ðoàn 1992), prosodic words in Kammu may take either a stop, a fricative, a nasal, a liquid, or a glide as coda (Svantesson 1983).In the extreme case, Lahu (Matisoff 2001, 2006) disallows codas altogether. However, there are in fact some characteristics with regards to codas that are shared by monosyllabic languages of SEA. First, these languages lack phonation-type contrast in coda stops (Matisoff 2001; Pittayaporn 2005a; Rhee 2003). Second, it is perhaps safe to say that these languages quite consistently disallow complex codas (Rhee 2003). Nevertheless, these charactieristics are independent of the monosyllabicism because they are found in disyllabic and sesquisyllabic languages as well. In sum, the canonical word shape in monosyllabic languages in SEA tends to be either CV(C) or CCV(C). That is, the prosodic word must be exactly one heavy syllable, but it is not clear what kind of a foot it is. When complex onsets are allowed, the second consonants almost always have relatively high sonority. Whether complex onsets are allowed or not depends on the individual languages. As for codas, there does not seem to be any correlation between monosyllabicism and possible final consonants. 2.4.2 Disyllabic languages Disyllabic languages refer to linguistic varieties in which the prosodic word predominantly consists of exactly two syllables. This type of languages is found only rarely in Mainland SEA but is widely attested in languages of Insular SEA. Examples of disyllabic languages in SEA include Acehnese (Durie 1985), Northern Roglai (Awơi-hathe et al. 1977), and Moken (Chantanakomes 1980; Pittayaporn 2005a). As in monosyllabic languages, both word maximality and word minimality must work together to keep each prosodic word exactly one metrical foot. For instance, Northern Roglai (Awơi-hathe et al. 1977), an Austronesian language of Central Vietnam, allows 56 only heavy monosyllables and disyllables consisting of a light syllable followed by a heavy one. Words that consist of only one light syllable, e.g. */bĭ/ or */mŭ/, do not exist in this language. In terms of foot typology, a bisyllabic foot can be either iambic or trochaic. Moken (Pittayaporn 2005a), Acehnese (Durie 1985), and Northern Roglai (Awơihathe et al. 1977; Lee 1966) are clear examples of iambic disyllabic languages. In these languages, the stressed syllable of a foot is always heavy, a fact which follows directly from Hayes’s typology in §2.2.2.314. Indonesian is an example of a disyllabic language with syllabic trochees (Cohn in progress). Although words can be longer than two syllables, an overwhelming majority of morphologically-simple words are bisyllabic. Interestingly, to my knowledge, no language with moraic trochees is attested in SEA. Also, note that these example languages all belong to the Austronesian language family. Like their monosyllabic counterparts, disyllabic languages differ from one another by whether they allow complex onsets or not. Northern Roglai is an example of a disyllabic language with iambic feet. Morphologically simple words in Northern Roglai are predominantly disyllabic, although monosyllabic words are also common. Some disyllables are /rawak/ ‘bridge’, /publɤj/ ‘to sell’ and /tijuh/ ‘seven’, contrasting with the following monosyllabic words /wak/ ‘blade for weeding’, /blɤj/ ‘to buy’, and /joh/ ‘to break’. According to my preliminary examination of the vocabulary, there is no restriction on what consonants can function as onsets of the two syllables. In other words, C1 does not limit the choice of C2 in disyllabic C1V.C2VC forms. Table 2-8 gives the inventory of northern Roglai complex onsets. 14 Hayes (1995: 82) acknowledges iambs consisting of two light syllables, i.e. even iambs, but states that they are rare and unstable. 57 Table 2-8 Complex onsets in Northern Roglai labial alveolar palatal velar C- pr- tr- cr- krˀC- -rC̬ - br- dr- ɟr- grC- pl- tl- klˀC- -l- ˀblC̬ - bl- dl- gl- Examples of Northern Roglai words with complex onset include /ɓlop/ ‘hold in mouth’, /trom/ ‘trunk’, and /kacreh/ ‘sarong, skirt’. This means that onsets in Roglai also obey SSP and have the sonority distance of at least 2. In other words, the second element of a cluster must be more sonorous than the first consonant. In particular, only complex onsets that consist of a stop followed by a liquid are allowed. In contrast to Roglai, Indonesian is a good example of a disyllabic language that by and large does not have complex onsets. The canonical shape of the Indonesian prosodic word is C1VC2.C3VC4. While only one consonant is allowed syllable-initially, there is no restriction on what consonants can function as onsets of the two syllables, i.e. C1 and C3. This lack of co-occurrence restriction of the two onsets is identical to the situation in Northern Roglai but is in sharp contrast with the situation in sesquisyllabic languages (see §2.4.3). As in monosyllabic languages, the set of permissible codas in each language seem unrelated to the disyllabicism. While Mishmi (Sastry 1984) only allows final stops and nasals, prosodic words in Indonesian may take a stop, a fricative, a nasal, a liquid, or a glide as coda (Rhee 2003). Moreover, disyllabic languages also share with monosyllabic languages the lack of phonation-type contrast in coda stops (Matisoff 2001; Pittayaporn 2005a; Rhee 2003) and the ban against complex codas (Rhee 2003). 58 These exact same characteristics are also found in monosyllabic and sesquisyllabic languages. In short, the canonical shape of the prosodic word in disyllabic languages is C(C)V(C).C(C)V(C). When complex onsets are allowed, the second consonants have relatively high sonority. However, the onsets of the two syllables do not constraint each other. Many disyllabic languages do allow monosyllabic words although they often require their prosodic word to be no smaller than a heavy syllable. 2.4.3 Sesquisyllabic languages Sesquisyllabic languages are languages whose prosodic words canonically consist of one syllable and a half. The term “sesquisyllable”, literally “one-and-a-half syllable”, was originally coined by Matisoff (1973) to refer to words that consist of an unstressed “minor” syllable followed by a fully stressed “major” syllable. The syllabicity of the minor syllable is carried by a neutral vowel or a syllabic consonant (Diffloth 1976: 232; Svantesson 1983: 27). Even though a priori the minor syllable can be located on either side of the major syllable, in the context of SEA the minor syllable always precedes the major syllable. Although Sloan (1988), Shaw (1994), Cho and King (2003), and I (Pittayaporn 2005c) have all argued that the minor syllable is a prosodically degenerate syllable of some kind, similar to the case of Polish clusters discussed in §2.2.2.2, it is likely that the so-called sesquisyllables differ structurally from one language to another. In this dissertation, I define a sesquisyllable as a prosodic word that consist of two syllables, the first of which is unstressed and lacks a phonological vowel. According to this definition, a sesquisyllable contrasts with a disyllable in that the first syllable of a disyllable does have a contrastive vowel. An important typological observation is that sesquisyllabic languages usually allow for monosyllabic words while monosyllabic languages do not allow 59 sesquisyllabic forms. This difference is clearly illustrated by comparing the sesquisyllabic Kammu and its monosyllabic relative Vietnamese. A good example of a sesquisyllabic language is Kammu discussed above in relation to the importance of the prosodic word. Other examples of sesquisyllabic languages include Burmese (Green 2005), Palaung (Kasisopa 2003), Tariang (L-Thongkum 2001), Chru (Jrang et al. 1983), and Jingpho (Maran 1978), to cite a few. With respect to the word size requirements, sesquisyllabic languages show effects of both word maximality and word mininallity so that well-formed prosodic words are exactly one metrical foot. For example, a Kammu prosodic word is either C(C)VV, C(C)V(V)C, C(C).C(C)VV or C(C).C(C)V(V)C as discussed in §2.2.1. Assuming that the minor syllable is light, a sesquisyllable must be considered a prototypical iamb. This means that sesquisyllabic languages require their prosodic words to be exactly one iambic foot: either a single heavy syllable or a light syllable followed by a heavy syllable. Iambic feet consisting of two light syllables, also known as even iambs do not appear to be allowed in these languages (Pittayaporn 2005c). Although the formal analysis of sesquisyllables may still leave room for discussion what sesquisyllables are formally, it is important to draw a distinction between sesquisyllables and disyllables (see §2.4.2). The crucial difference between the two types of prosodic word is that the onsets of the major and minor syllables interact very closely. While the onsets of the two syllables in disyllables are usually independent from each other, the onsets of the minor and major syllables in sesquisyllabic words may pose restrictions on each other. The Kammu language exemplifies such interaction very clearly. Prosodic words whose minor syllable and major syllable onsets are identical, i.e. */p.p-/, */t.t-/, */s.s-/ etc., are not attested in Kammu as described by Svantessoon (1983). Second, sesquisyllabic words often violate SSP or show lower sonority distance than 2 (see §2.2.2.2). Therefore, the 60 minor syllable and the major syllable onsets together will be referred to in this study as “sesquisyllabic clusters”. This term is used in opposition to “tautosyllabic clusters”, which refers to true clusters that function as the onset of a single syllable. Sesquisyllables also differ quite markedly from prototypical monosyllables. While complex onsets of monosyllables, i.e. monosyllables with unmarked initial clusters, obey SSP and show sonority distance of at least 2, complex onsets of sesquisyllables do not necessarily follow the generalizations. This distinction is shown in Table 2-9. Table 2-9: Onsets and syllables in Kammu (from Pittayaporn 2005c) Initials monosyllabic sesquisyllabic obstruent+obstruent   obstruent+nasal   obstruent+liquid   We see that when a consonant sequence in Kammu violates SSP or has relatively low sonority distance among its members, it is parsed as sesquisyllabic. For instance, a sequence of two obstruents are always realized as sesquisyllabic, e.g /p.téʔ/ ‘earth’, never */ptéʔ/. Similarly, a sequence of a liquid followed by an obstruent is also parsed as a sesquisyllabic cluster, e.g. /l.cɔ̀:t/ ‘to become snotty’, never */lcɔ̀:t/. On the other hand, if the consonant sequence obeys the SSP, it may be realized either as a sesquisyllabic or tautosyllabic cluster. Kammu in particular contrast /kl-/ as in /kló:k/ ‘bamboo bowl’ and /k.l-/ as in /k.ló:k/ ‘slit drum’. Independent evidence in Kammu for this contrast comes from tone patterning. Furthermore, in words whose major syllables have onset clusters, those onset clusters must respect the sonority hierarchy. For example, /s.kró:/ ‘to suddenly bend down head’ has an initial sesquisyllabic 61 cluster, which in turn consists of an initial /s-/ and a well-formed tautosyllabic cluster /kr-/. These generalizations are based only on Kammu, but to my knowledge they seem to hold for other sesquisyllabic languages as well. They play a crucial role in the reconstruction of PT complex onsets in §2.3.2 and Chapter 4. In addition, many sesquisyllabic languages allow minor syllables that are more complex than a single consonant. For example, Jahai (Burenhult 2005), a Mon-Khmer language of Malaysia, has an abundance of sesquisyllabic words whose minor syllables consists of two consonants, e.g. /sl.tuh/ ‘to attack’, /cn.rɔs/ ‘nail’, /ʔn.teŋ/ ‘ear’. However, the second consonant has to be a sonorant and must agree in place of articulation with the onset of the major syllable if it is a nasal. Similarly, Kammu also allows for minor syllables that consist of two consonants but in addition to sonorants it also permits obstruents as the second consonant, cf. /rt̀ .ʔàt/ ‘plug’, /pc̀.rà:c/ ‘mesentry’. However, in Kammu these obstruents must be identical to the coda of the major syllable (Svantesson 1983: 28-29)15. As in the other types of languages, the set of permissible codas in each language seem unrelated to the fact that they are sesquisyllabic. While Buyang only allows final stops and nasals (Li 1999), the coda of the major syllable of prosodic words in Nyah Kur may take a stop, a fricative, a nasal, a liquid, or a glide as coda (Diffloth 1984). Furthermore, sesquisyllabic languages also share with monosyllabic and disyllabic languages the lack of phonation-type contrast in coda stops (Matisoff 2001; Pittayaporn 2005a; Rhee 2003) and the ban against complex codas (Rhee 2003). These restrictions on codas are clearly a Southeast Asian phonological trait (Matisoff 2001; Rhee 2003). 15 This type of Kammu minor syllables is always morphologically complex, created either by prefixation or infixation (Svantesson 1983: 29). 62 In sum, the canonical shape of a prosodic word in sesquisyllabic languages is exactly one iambic foot, and may have one of the following shapes: C(C)VV, C(C)VC, C(C).C(C)VV, or C(C).C(C)V(V)C. While tautosyllabic clusters in sesquisyllabic languages obey SSP, and have the sonority distance of 2, sesquisyllabic clusters may or may not. 2.5 PT as a sesquisyllabic language Having argued in §2.3.2 that the sesquisyllabic hypothesis is the best hypothesis for the reconstruction of PT, in this section I characterize PT as a sesquisyllabic language, very similar typologically to Kammu discussed in §2.4.3. The PT prosodic word can be either a monosyllable or a sesquisyllable, as argued for in §2.3. Moreover, comparative data discussed in detail in the following chapters permits us to view the PT word as having the following characteristics. First, each prosodic word contained exactly one iambic foot headed by a heavy syllable. Second, phonation-type and place contrasts in the onset were neutralized in the coda position. Last, each prosodic word is specified for one lexical tone. This section provides an explicit characterization of the PT prosodic word. Diachronically speaking, this stage is one step away from the monosyllabicism observed in modern languages along the evolutionary path taken by Chamic languages and others outlined above. In synchronic terms, the PT prosodic word was exactly one foot. In other words, word minimality and maximality both required that a prosodic word be a well-formed metrical foot, as in other sesquisyllabic languages. Furthermore, the metrical foot in PT was iambic as in sesquisyllabic languages (see §2.4.3). The prominent syllable of the PT prosodic word must have been the final syllable, i.e. the major syllable. Monosyllabic words were also permissible in PT 63 because a heavy syllable can by itself be an iamb. Possible shapes of PT prosodic words according to the current proposal are given in Table 2-10. Table 2-10 Canonical shape of PT prosodic word Monosyllable Sesquisyllable Open syllable *C(C)(C)VT C(C)(C)VVT *C(C).C(C)(C)VT C(C).C(C)(C)VVT Closed syllable C(C)(C)V(V)CT C(C).C(C)(C)V(V)CT C = consonant, V = vowel, T = tone, * = not a possible word The PT prosodic word can be divided into three parts for the purposes of reconstruction: 1) onset, 2) rime, and 3) tone. The onset consists of any consonant that precedes the vocalic nucleus including the minor syllable. That is, it includes both the tautosyllabic and sesquisyllabic onsets as characterized above. The rime includes both the vocalic nucleus and the final consonant, if any. Lastly, the tone is the suprasegmental contrast linked to the major syllable; the minor syllable does not carry tone. The PT prosodic word contained exactly one iambic foot. Comparative data discussed in §5.3 indicates that PT prosodic words ended in either open or closed syllables as schematized in Table 2-10. The only prohibition is against open syllables with one short vowel. In other words, only a heavy syllable can stand as a prosodic word by itself in the case of monosyllables, or function as the major syllable in the case of sesquisyllables. This is a clear manifestation of the minimality and maximality requirements. In terms of foot typology, this means that even iambs, i.e. iambic feet consisting of two light syllables, were not allowed. In the following subsections, I will outline the phonotactics of each of the three components according to the current proposal. An explicit characterization of the PT prosodic word is given in Figure 2-3. 64 PrW ! φ ! σ gy μμ !! C(C)(C) V V T PrW ! φ ! σ gy μμ !! C(C)(C)V(V)C T a. open monosyllable b. closed monosyllables Examples: *ʰma:A ‘dog’, *pi:A ‘year’, *ɗa:B ‘to scold’ Examples: *ʰmatD ‘flea’, *ɓlo:kD ‘flower’, *gapD ‘narrow’ PrW ! φ ty σσ ! gy μ μμ ! !! C (C) C(C)(C)V V T PrW ! φ ty σσ ! gy μ μμ ! !! (C) C C(C)(C)V(V)C T c. open sequisyllable d. closed sesquisyllables Examples: *p.ta:A ‘eye’, *k.ro:A ‘to beg’, *C̥ .ba:B ‘shoulder’ Examples: *p.ta:jA ‘to die’, *k.temA ‘full’, *gm.ra:C ‘orphan’ Figure 2-3 Phonological representation of PT prosodic word 2.5.1.1 Onset The PT prosodic word may have simple or complex onsets. A simple onset may be any of the consonants in the PT sound inventory16. PT had a three-way phonation type contrast among stops—voiceless, implosive, and voiced (see §3.5). 16 The proposed inventory of PT consonants is given in Table 3-1. 65 Moreover, it contrasted altogether six places of articulation—labial, alveolar, palatal, velar, uvular, and glottal (see §3.3). Complex onsets may be either tautosyllabic or sesquisyllabic. A tautosyllabic cluster is a sequence of consonants that functions as onset of a single syllable as illustrated in Figure 2-3a and Figure 2-3b. A tautosyllabic cluster in PT may consist of up to three consonants and must be canonical clusters. In other words, they respect SSP and show a minimum sonority distance of 2. Exceptions are tautosyllabic clustes containing a liquid followed by /-w-/, e.g. *lw-, *rw-, to cite a few. In these cases, the sonority distance between the liquid and the glide is only 1 on the sonority scale. Examples of permissible PT tautosyllabic clusters include *kr-, *kl,*kw-, and *krw- (see §4.2). In contrast, a sesquisyllabic cluster is a sequence of consonants that makes up the minor syllable plus the major syllable onset. A PT sesquisyllabic cluster in theory can have up to two consonants in the minor syllable, and up to three consonants in the major syllables, even though no sesquisyllabic clusters with five consonants is posited in this study. Examples of sesquisyllabic clusters are *k.t-, *p.k-, *k.r-, and *gṃ.r(see §4.4). Crucially, in PT a contrast between *kt- and *k.t- is not possible but tautosyllabic *kr- did contrast with sesquisyllabic *k.r-. 2.5.1.2 Rime The PT rime consists of two sub-parts—the vocalic nucleus and the coda. As represented in Figure 2-3, only the major syllable of a sesquisyllabic word has a vocalic nucleus. Therefore, the rime of a sesquisyllable refers specifically to the rime of the major syllable. PT vocalism shows a three-way vowel height contrast as well as a vowel length distinction (see §5.3). Any vowel in the PT phoneme inventory can occupy the nucleus of the syllable. However, the size requirement dictates that each prosodic word be exactly one iambic foot as represented in Figure 2-3. One 66 consequence is that open syllables with short vowels are banned. Open monosyllables with short vowels were banned because they are smaller than a foot. Similarly, sesquisyllabic forms with open major syllables ending in short vowels were not allowed because light syllables cannot head an iambic foot. Examples of words allowed by the PT phonotactics are *ka:A, *katD, and *ka:tD, but not *kaA. The second part of the rime is the coda. For each syllable, one coda consonant was allowed as represented in Figure 2-3 Not all PT consonants can occupy the final position of a prosodic word as phonation-type and place contrasts were neutralized in this position. Only the voiceless stops, the nasals, the lateral liquid, and the glides could function as codas. Similarly, not all places of articulation are found in the coda position. Of the six series, only labial, alveolar, palatal, and velar consonants can occupy the final position of a syllable (see §5.5). 2.5.1.3 Tone Each PT prosodic word contained one lexical tone as represented in Figure 2-3. In sesquisyllabic words, only the major syllables carried distinctive lexical tones. A syllable ending in a vowel or a sonorant, “non-checked syllable”, could have either tone *A, *B, or *C. In contrast, a “checked syllable”, or stop-final syllable, is conventionally said to have had only one tone *D. From the point of view of phonological contrast, we can say that PT checked syllables did not contrast for tone. The distinction between *D on one hand, and *A, *B, and *C, on the other, was therefore one of syllable structure (§6.2). The three tones on non-checked syllables are tentatively posited as a modal tone with mid level contour, a creaky tone with low rising contour and relatively long vowel duration, and a finally glottalized tone with a high falling contour and short vowel duration (see §6.5). 67 2.6 Summary In this chapter, I discussed the structure of the prosodic word in PT. I first discussed the importance of the prosodic word in the reconstruction of PT. I reviewed the monosyllabic, sesquisyllabic, and disyllabic views of PT and showed that the sesquisyllabic view can account for the range of comparative data as well as offer a plausible scenario of how PT developed into daughter languages. I have thus proposed that PT was a sesquisyllabic language whose prosodic word can be either a monosyllable or sesquisyllable. After providing a brief typological survey of the canonical shape of the prosodic word in SEA languages, I outlined the phonotactics of PT onsets, rimes, and tones. In contrast to earlier reconstruction, this dissertation addresses the issue of the PT prosodic word explicitly. The canonical shape of the PT prosodic word that has been established will be the framework for the segmental reconstruction. Detailed discussion of each part together with their reconstruction will be given in subsequent chapters. 68 CHAPTER 3 PROTO-TAI SIMPLE ONSETS 3.1 Introduction Among the three components of the PT prosodic word, the onsets are probably the best understood. However, many issues await solutions. The best-known problem regarding the onsets of PT is the so-called “voicing alternation,” which involves discrepancies in the voicing of the onset initial between different groups of dialects as discussed in §1.3.2. In addition, one issue that has usually been overlooked is the number of places of articulation. In contrast to the conventional view that PT had labial, alveolar, palatal, velar, and glottal series, Haudricourt (1952), and Ostapirat (2007) posit a distinct series of uvular consonants, which seems to be preserved in PSWT (Pittayaporn 2008b, to appear-b). Moreover, I have elsewhere discussed gaps in the inventory of PSWT initial consonant clusters that need to be accounted for at the PT level (Pittayaporn 2008b, to appear-b). In this chapter, I propose an inventory of simple PT onsets that effectively accounts for these outstanding problems. As a point of departure, I present the inventories of simple and complex onsets of PT and outline the phonotactics of the PT onsets. Then, I discuss three aspects of the present system: 1) the reconstruction of the uvular series, 2) the treatment of the so-called “voicing alternation,” and 3) the lack of contrastive aspiration. Lastly, I present the reconstructions of each of the PT simple onsets. 3.2 Inventory of PT simple onsets Each PT syllable had an onset. I propose that there were 36 consonants in PT, all of which can occur alone as simple onsets. The proposed PT inventory is given in 69 Table 3-1. This section gives an overview of the onset system of PT, highlighting notable characteristics involving the manners and places of articulation as well as the phonation types. In the table, three consonants are put in parentheses because their reconstruction is tentative. Table 3-1 PT consonant inventory stops voiceless voiced glottalized fricatives voiceless voiced nasals voiceless voiced liquids and glides voiceless voiced labial alveolar palatal *p *t *c *b *d *ɟ *ɓ *ɗ *ˀj *s (*ɕ) *z (*ʑ) *ʰm *ʰn *ʰɲ *m *n *ʰw *ʰr *ʰl *w *r *l *ɲ velar *k *g *x *ɣ (*ʰŋ) *ŋ uvular *q *ɢ *χ glottal *ʔ *h In this proposal, PT consonants can be divided into four groups according to their manner of articulation: 1) stops, 2) fricatives, 3) nasals, and 4) liquids and glides. Unlike Li who views the two palatal obstruents as affricates (Li 1977: 164), I consider them as stops phonologically. The choice is based mainly on the fact that these palatal consonants behave like stops in other places of articulation. First, the voiceless *c can occur as a coda, like *-p, *-t and *-k, as discussed in Chapter 5. Second, the voiced *ɟ always follow the same devoicing pattern as *b-, *d- and *g-. In languages where devoicing yields aspiration, *ɟ-, like *b-, *d-, and *g-, is reflected as an aspirated stop 70 /cʰ-/ or its derivative /s-/, e.g. Siamese /cha:jA2/ ‘male’ and Lao /sa:jA2/ (from earlier *cʰa:jA2). Similarly, these voiced obstruents are reflected as plain in languages that do not aspirate their reflexes of PT voiced *b-, *d-, *ɟ- and *g-, e.g. Black Tai /ca:jA2/ and Lungchow /ca:jA2/. Lastly, treating *c and *ɟ as stops makes the system economical and symmetrical. If they were considered affricates, a separate manner of articulation must be posited only for the two of them. Treating *c and *ɟ as stops also makes the palatal series very similar to other places of articulation. However, this is not to argue that PT *c and *ɟ were not affricated phonetically. As for places of articulation, six different categories are posited for the proposed PT consonants: 1) labial, 2) alveolar, 3) palatal, 4) velar, 5) uvular, and 6) glottal. This reconstruction differs from most proposals (Haudricourt 1948; Li 1977; Nishida 1954; Sarawit 1973) by positing a distinct series of uvular consonants, but agrees with Haudricourt (1952)’s and Ostapirat (2007)’s proposals that uvular consonants be reconstructed for PT. The reconstruction of uvular consonants will be discussed in §3.3. Note that uvular and glottal places of articulation are defective in that they lack many consonant types found in the other places of articulation. Most notable is the lack of sonorants, which seems to be typologically common for places of articulation further back in the vocal tract. As for phonation type, there was a three-way contrast in PT: 1) voiceless, 2) implosive/glottalized, and 3) voiced. When occurring as simple onsets, PT voiceless stops are reconstructed as unaspirated because they are reflected as unaspirated in all modern languages. In contrast, PT voiceless sonorants, except for *ˀj- are reconstructed as phonetically pre-aspirated because, as far as tonal development is concerned, they behave like aspirated stops in the modern varieties that have both plain and aspirated stops (Gedney 1972a; Li 1977). Moreover, the pre-aspiration is attested in loanwords from Tai to Austroasiatic languages, e.g. Red Palaung /ʰlа:n/ 71 ‘grandchild’ and /ʰnoŋ/ ‘lake’ ultimately from PT *ʰla:nA and *ʰno:ŋA respectively (Pittayaporn to appear-a). In addition, the implosives stops are viewed as half-voiced and half-voiceless. Specifically, they behave like voiceless consonants in terms of tonal development when occurred with certain tonal categories but like voiced consonant when occurred with others (Gedney 1972a). The implosives *ɓ- and *ɗ- correspond to *ˀb- and *ˀd- in earlier constructions. Although the choice between implosive or glottalized stops does not have any implication to the rest of the system, reconstructing these sounds as implosives is preferable from a typological point of view. Specifically, cross-linguistically implosive /ɓ/ and /ɗ/ are more common than their counterparts further back (Croft 1990: 159; Maddieson 1984b: 111-114), However, these PT implosive stops pattern with *ʔ- and *ˀj- in relation to tonal splits and mergers (see §6.2). For convenience, the series that consists of the implosives and the two glottalized consonants is referred to here as glottalized. 3.3 A richer dorsal series Among the five contrastive series of consonants proposed in Li’s reconstruction, the velar series is the most problematic one. There are many inexplicable split reflexes for the many simple and complex velar onsets that he reconstructs. For example, each of Li’s *kʰ- and *x- may be reflected as either /kʰ-/ or /x-/ in White Tai, but the conditioning environment for the split cannot be identified. Another example is Li’s cluster *kr- which is reflected irregularly across dialects. Based on data from newly-described SWT languages, I have argued elsewhere that PSWT had two contrastive uvular consonants *q- and *χ- (Pittayaporn 2008b, to appear-b). At the PT level, Haudricourt (1952) proposes that there was one distinctive uvular stop *q- while Ostapirat (2007) believes that both the voiceless *q- and the 72 voiced *ɢ- existed. Comparing the SWT data with data from NT dialects, I posit three uvular consonants for PT: *q-, *χ-, and *ɢ-. Table 3-2 shows clearly that the velar series alone cannot account for the range of correspondences that involve dorsal onsets. Table 3-2 Voiceless dorsal correspondence sets Li Gloss 1 *x-, *kʰ- ‘rice’ ‘to open’ ‘guest’ ‘white’ 2 ‘galangal’ *x- ‘green’ ‘to cross’ 3 ‘to enter’ ‘ginger’ 4 *kʰ- ‘to kill’ ‘torn’ ‘arm’ PT PSWT - *kʰ*x- *x*χ- *χ*q- *q- SWT White Kapong Tai kʰawC1 kʰajA1 kʰɛʔDL1 kʰawC1 kʰajA1 kʰɛʔDL1 xa:wA1 xa:B1 xɛwA1 kʰa:wA1 kʰa:B1 kʰe:wA1 xa:mC1 xawC1 xiŋA1 ha:mC1 hawC1 hi:ŋA1 xa:C1 xa:tD1 xɛnA1 ha:C1 ha:tD1 hɛ:nA1 NT Yay Wuming hawC2 ha:jA1 hekDL1 ha:wA1 hewA1 ha:mC1 hawC1 hiŋA1 ka:C1 ka:tD1 cenA1 xawC2 ka:jA1 hekDS1 ha:wA1 haA1 hewA1 ha:mC1 hawC1 ka:C1 ka:tD1 kenA1 The etyma in set 1 must be reconstructed as *kʰ- at the PSWT level. However, this PSWT aspirated velar stop cannot be projected back to PT because the etyma in this set either show voicing alternation (see §3.4), or are clearly relatively recent loans. The etymon for ‘rice’ along with many others have initials that go back to original voiceless onsets in SWT and CT but original voiced onsets in NT. The forms for ‘to open’ and ‘guest’ must be considered later borrowings from Chinese 開 kāi and 客 kè because they are irregular in NT. In addition to the aberrant initial /k-/ in Wuming, the vowels are unexpectedly long in ‘to open’ in both Yay and Wuming. The form for 73 ‘guest’ in Wuming is also irregular as it shows DS1 tone rather than the expected DL1. Although we can safely follow Li (1977: 207-214) in reconstructing /x-/ for etyma in set 2, Kapong forms in set 3 point to a different proto-phoneme. For both sets, White Tai, Yay, and Wuming show /x-/, /h-/, and /h-/ respectively. However, the Kapong dialect of Phu Thai, spoken in northeastern Thailand, shows two different reflexes for the two sets of etyma—/kʰ-/ for set 2, and /h-/ for set 3. To account for Kapong /h-/, which is produced further back in the vocal tract, I posit a uvular *χ- for etyma having the correspondence pattern of ‘to cross’, ‘to enter’, and ‘ginger’. PT *xcan be kept for ‘white’, ‘galangal’, and ‘green’. Although one may argue that the split reflex in Kapong is an innovation, neither a segmental nor a tonal conditioning environment can be identified. For example, the variation in vowels in these forms precludes positing a conditioned backing of *x- in Kapong. The data in set 4 shows yet another pattern of correspondence. While the SWT dialects White Tai and Kapong have fricative reflexes /x-/ and /h-/ respectively, the NT dialects Yay and Wuming both show a stop /k-/. While one may be tempted to posit PT *kʰ- to account for this disprepancy, data from the CT dialects of Long’an and Yongnan in central Guangxi rule out the possibility. In these two dialects, aspiration is contrastive, e.g. Yongnan /kʰatDS1/ ‘section’ vs. /katDS2/ ‘tight’. Forms that are reconstructed with aspirated stop in Li’s system mostly show aspirated onsets as illustrated by Yongnan /pʰla:A1/ ‘stone mountain’, /pʰlakDS1/ ‘vegetable’, /tʰrənA1/ ‘stone’, and /tʰa:nB1/ ‘charcoal’ for Li’s *pʰraA, *pʰrəkD, *tʰrinA, and *tʰanB respectively. If the etyma in set 4 had had *kʰ- as onsets in PT, we would expect to see aspirated reflexes in Long’an and Yongnan. However, the two dialects regularly show unaspirated reflexes in etyma belonging to set 4 as exemplified by Yongnan /ka:C1/ ‘to 74 kill’, /ka:tD1/ ‘to tear’, and /ke:nA1/ ‘arm’. These two crucial dialects are therefore strong evidence for a lack of aspiration in the PT onsets of etyma belonging in set 4. Therefore, I posit an unaspirated uvular *q- for this correspondence. Moreover, positing PT *q- for this set of etyma is also preferable from a typological perspective. That is, it is very unlikely for a language to have a fricative but lack the corresponding stop, except for the glottal series. The development of these voiceless initial dorsal consonants in White Thai, Kapong, and Yay is schematized in Figure 3-1. PT White Tai PT Kapong PT Yay *k- k*x- *k- *k- *kk- *x- *kh- *x- *q- x- *q- *q- *χ- *χ- *h- *χ- h- *h- h- *h- *h- Figure 3-1 Development of dorsal onsets in White Thai, Kapong, and Yay Not only is there evidence for the voiceless uvular set, data from modern dialects also point to voiced uvular onsets that contrasted with voiced velar onset in PT. Ostapirat (2007) shows that correspondence sets that Li assigns to *ɣ- in fact consist of two correspondence patterns. For one pattern, White Tai shows /x-/ regularly, pointing to PSWT *ɣ-. For the other pattern, the regular White Tai reflex is /k-/ suggesting PSWT *g-. He argues that there is no evidence for treating the two patterns as reflexes of two variants of the same sound and proposes that Li’s *ɣ- must in fact be reconstructed as two different phonemes, *ɣ- and *ɢ-. Data from a wider range of dialects given in Table 3-3 support this position. 75 Table 3-3 Voiced dorsal correspondence sets Li Gloss PT ‘pair’ 1 *g- ‘swollen’ *g- ‘to chew’ ‘carry pole’ 2 ‘chin’ *ɢ- *ɣ- ‘human’ ‘night’ 3 ‘thatch grass’ *ɣ- ‘shin’ PSWT *g*ɣ- SWT White Tai ku:B2 kaɰB2 kɛwC2 ka:nA2 ka:ŋA2 kunA2 xɔA2 xa:A2 xɛŋB2 CT Leiping Debao kʰu:B2 kʰaɰB2 kʰe:wC2 kʰa:nA2 kʰa:ŋA2 kʰɤnA2 ho:A2 ha:B2 he:ŋB2 kowB2 kɔjB2 kewC2 ka:nA2 ka:ŋA2 kɔnA2 vo:A2 ja:A2 - NT Yay ku:B2 - cewC2 ha:nA2 ha:ŋA2 hunA2 hɔA2 haB2 heŋB2 The etyma in set 1 clearly go back to voiced velar stop *g-. This PT onset is reflected regularly as a stop in White Tai, Leiping, Debao, and Yay. Note that original voiced stops regularly yield aspirated stops in Leiping, cf. /pʰi:A2/ ‘fat’ from *bwi:A, and /tʰo:ŋC2/ ‘stomach’ from *dwu:ŋA. The etyma in set 2 show initial /k-/ and /kʰ-/ in SWT and CT languages, indicating that the original onset was a stop. However, the regular Yay reflex for this set is /h-/, indicating that that stop cannot be *g-. Therefore, I follow Ostapirat (2007) in positing PT *ɢ- for these items. Lastly, the etyma in set 3 are reflected as /x-/ in White Tai and /h-/ in Leiping and Yay, indicating that Li was correct to list them under *ɣ-. Note that the Debao reflex of PT *ɣ- is determined by the following segment. It is reflected as /v-/ before labial sounds and as /j-/ elsewhere. Interestingly, the corresponding Sui forms for ‘thatch grass’ /ja:/ and ‘dike between fields’ /jan/ cited as examples of PT *ɣ- in Ostapirat (2007) also have sonorant onsets. The development of the voiced dorsal onsets in White Tai, Debao, and Yay is schematized in Figure 3-2. 76 PT White Tai PT Debao PT Yay *gk- *ɢ*ɣ- x- *gk- *ɢ*ɣ- v- *g- k*ɢ- h*ɣ- j- Figure 3-2 Development of dorsal onsets in White Thai, Debao, and Yay Although none of the attested daughter languages have distinctive uvular consonants, I have provided solid comparative evidence for a distinct series of uvular consonants in PT. For this place of articulation, at least three phonemes must be posited: the voiceless stop *q-, the voiceless fricative *χ-, and the voiced stop *ɢ-. Unlike its velar counterpart, the uvular place of articulation seems to be defective as it lacks sonorants altogether. Given that the acoustic structures of velar and uvular consonants are very similar (Ladefoged & Maddieson 1996: 36-37), we can view the lack of uvular sonorants as neutralization of dorsal consonants. Refer to §3.6.1.5, §3.6.3.5, and §3.6.4.4 for reflexes of individual etyma in different languages. 3.4 “Voicing alternation” revisited One of the major problems in the field of Comparative Tai is the so-called “voicing alternation,” which refers to an apparent discrepancy regarding the voicing of onsets in different dialects. This discrepancy is generally described as a disagreement between SWT and CT on one hand, and NT on the other. For approximately 50 etyma found widely among Tai languages, tonal reflexes in SWT and CT points to earlier voiceless onsets while their NT counterparts indicate original voiced onsets (Gedney 1989a). For illustration, some cases of voicing alternation are given in Table 3-4. 77 Table 3-4 Examples of voicing alternation Gloss ‘male’ ‘to carry’ ‘to ride’ ‘ripe, cooked’ Li Siamese Lungchow Po-ai *pʰ- pʰu:C1 pʰu:C1 pu:C2 *tʰ- tʰɯ:A1 tʰɯ:A1 tɯ:A2 *kʰ- kʰi:B1 kʰwi:B1 kɯjB2 *s- sukDS1 ɬukDS1 ɕukDS2 The etymon ‘male’ is reflected in Siamese and Lungchow as /pʰ-/ with tone C1 indicating earlier voiceless *pʰ- but reflected in Po-ai as /p-/ with tone C2 pointing to original voiced *b-. Similarly, ‘to carry’ and ‘to ride’ are reflected in the two non-NT dialects as /tʰ-/ and /kʰ-/ with tones in the first series, indicating *tʰ- and *kʰ- at an earlier stage. For these same two etyma, Po-ai shows /t-/ and /k-/ with tones belonging to the second series, which are regular reflexes of voiced *d- and *g-. As for ‘ripe, cooked’, the /s-/ and /ɬ-/ with tone DS1 in Siamese and Lungchow point to earlier *s-, but /ɕ-/ with tone DS2 in Po-ai indicates an original voiced initial17. These cases of etyma showing voicing alternation have remained a serious problem in the reconstruction of PT onsets. Li (1977: 36-39) suggests that the voicing alternation was perhaps a derivational morphological process. He points out that there are doublets in Siamese that seem to show both the voiced and voiceless reflex. He cites Siamese pairs like /camA1/ ‘to confine, retain’ vs. /cʰamA2/ ‘food susceptible of storage’, and /kʰiəwC1/ ‘fang’ vs. / kʰiəwC2/ ‘to chew’. Gedney (1989a), however, argues convincingly that the so-called voicing alternation cannot be morphological in nature. First, no specific 17 NT points to *ɟ- rather than *z-. 78 morphological or syntactic function can be assocaited with these the alternating etyma. Second, it is rarely the case that a given modern language shows both the voiced and the voiceless variants. He also pointed out that it is rarely the case that both members of the pairs of Siamese words cited by Li are reconstructible at the PT level. In addition to Gedney’s arguments, it should also be pointed out that even when both members of the alleged pairs go back to PT, often the similarity between the reconstructed forms is not as great as the similarity between their corresponding Siamese forms. In such cases, it is doubtful I speculate that voicing alternationthat the two members of the alleged pairs actually go back to a single proto-form. For example, among Li’s Siamese doublets /kʰiəwC1/ ‘fang’ must be reconstructed with a uvular onset and a monophthong while /kʰiəwC2/ ‘to chew’ goes back to a velar onset and a diphthong: PT *χe:wC and *giəwC respectively. In cases such as these, there is no reason to posit a voicing alternation because the words were completely distinct at the PT level. Therefore, facts like these cast doubt on the claim that the observed pattern of voicing alternation has a morphological origin. In addition to Li (1977), a few proposals have been put forward. Ferlus (1990) posits sesquisyllabic forms to account for this voicing alternation, without providing any argument18. Gedney (1989a) in contrast, proposes that etyma showing the voicing alternation had a phonation type different from the four assumed in the conventional reconstruction. He speculates that this additional series might have been voiced and aspirated. Li (1989)’s later proposal closely resembles this position; the difference is largely notational. Lastly, Thurgood (2002a; 2007) proposes that forms showing the voicing alternation are post-PT borrowings, arguing that these etyma show internal 18 In the cited papers, he writes prenasalized stops as PT onsets of these voicing alternation items. However, he explains in a personal communication that the prenasalized stops are notational formulas for sesquisyllables of the type C.CVC or CN.CVC, where N represents any nasal. 79 irregularities and that many of them occur in other language families. Although a closer look at data from a wider range of Tai varieties seems to support Thurgood’s position, not all cases of voicing alternation can be explained away as loans. I conclude that voicing alternation items that cannot be treated as borrowings are native etyma that go back to PT sesquisyllables as Ferlus suggests. Examine the first set of data, given in Table 3-5. Table 3-5 Voicing alternation items that show additional irregularities Gloss ‘chopsticks’19 ‘log’ ‘to rake’ ‘to cross the arms’ ‘pole’ ‘boil, ulcer’ ‘to sharpen’ ‘eggplant’ ‘cloth’ ‘person’ ‘late in the morning’ Li Siamese *tʰ*kʰ- kʰɔ:nA1 *f- fɯəA1 *kʰ*s- sawA1 *f- fi:A1 *f- fonA1 *kʰ- kʰɯəA1 *pʰ- pʰa:C1 *pʰ- pʰu:C1 *s- sa:jA1 Lungchow pʰɯ:A1 pʰɤnA1 pu:C1 ɬa:jA1 Po-ai tɯ:B2 ku:nB2 fɯ:A1 ce:wB2 ɬa:wA2 pajA2 kɯəA2 pɯ:B2 puC2 kwa:jA2 Among the etyma showing voicing alternation, many show other irregularities in addition to the initial consonants. Thurgood (2002a; 2007) shows convincingly that roughly a quarter of the etyma listed above should be eliminated. For example, ‘chopsticks’ points to *u: in SWT and CT but *ɯ: in NT. This irregularity in the 19 This etymon is missing in Siamese and Lungchow. For SWT, see White Tai /tʰu:B1/. For CT, see Lungming /tʰowB1/. 80 vowel suggests that these modern forms for ‘chopsticks’ are later borrowings. The etymon ‘log’ cannot be reconstructed for PT because the modern forms do not agree in tone and show various other minor problems. This suggests that they might not be related at all. The etymon ‘eggplant’ is reflected with an /e:/ in Saek instead of the expected /uə/. The SWT and NT forms for ‘cloth’ also disagree with respect to tones suggesting that they may not be related. Another example is ‘late in the morning’, which shows irregularities in vowel length, and tone, as well as onsets. The NT form, which suggests earlier *gw-, may not be related at all. Refer to Thurgood (2007: 246248) for discussions of individual cases. Thurgood also points out that many etyma were borrowings from outside of Kra-Dai. Although the etymologies he proposed for many cases are not convincing20, it is clear that many of the voicing-alternation items are loans from other language families. The most important donor is of course Chinese. Table 3-6 presents etyma that have clear Chinese origins. 20 For example, he erroneously links the Tai etymon for ‘ear’ (*k.rwɯ: in SWT and CT) to Chinese 耳 ěr (MC ńźɨB < OC nəʔ) and Proto-Tibeto-Burman *g-na. Another example is the Tai etymon for ‘compartment’ (*suəmC in CT and SWT), which he connects to PAN *ɣumaq ‘house’. In this case, the Tai and Austronesian forms have very little in common. Their meanings also do not match well. 81 Table 3-6 Voicing alternation items that have clear Chinese origins21 Gloss ‘bean’ ‘bowl’ ‘young male animal’ ‘chopsticks’22 ‘sugar’23 ‘pond’ ‘row’24 ‘to ride’ ‘eggplant’ ‘to dig’25 ‘cooked, ripe’ ‘ten’ ‘enemy, war’26 ‘right (hand)’ Li Siamese Lungchow Po-ai *tʰ- tʰuəB1 tʰu:B1 tu:B2 *tʰ- tʰuəjC1 tʰu:jC1 tu:jC2 *tʰ- tʰɯkDS1 tɤkDS2 takDS2 *tʰ- tɯ:B2 *tʰ- tʰɤ:ŋA1 *tʰ- tʰumA1 tamA2 *tʰ- tʰɛwA1 te:wA2 te:wA2 *kʰ- kʰi:B1 kʰwi:B1 kɯ:jB2 *kʰ- kʰɯəA1 kɯ:A2 *kʰ- kʰut D1 kutDS2 hutDS2 *s*s*s*kʰw- sukDS1 sipDS1 sɯk DS1 kʰwa:A1 ɬukD1 ɬipD1 ɬa:A1 sukD2 sipD2 kwaA2 MC 荳 dəuC 斗 təwB 特 dək 箸 ḍjwoC 糖 daŋ 潭 dậm 條 diew 騎 gje 茄 gj 堀 kʰwət, or 掘 gjwət 熟 źjuk 十 źjəp 賊 dzək 右 jəuB The etyma in the table above all clearly have Chinese origins. Most items follow the regular tonal correspondences for shared Sino-Tai vocabularies (see §6.2), but many show unexpected tonal categories. For example, modern Tai forms for ‘to 21 The etymologies given in this column are from Thurgood, unless otherwise indicated. 22 For SWT, see White Tai, Black Tai, and Shan /tʰuB1/. For CT, see Leiping /tʰu:B1/, Lungming /tʰowB1/, Ningming /tʰaɰB1/, among others. 23 The MC form is from Pulleyblank (1991). 24 Author’s etymology 25 Author’s etymology 26 The etymology is my own. For CT, see Western Nung /sɤkDS1/, Ningming /sakDS2/, and Qinzhou /ɕakDS2/, etc. For NT, see Yay /sakDS2/, Wuming /ɕakDS2/, Yongbei /tsakDS2/, etc. The forms in Ningming, Qinzhou, Wuming, Yongbei, and other NT dialects, all means ‘to steal’. Also see no. 685 in Appendix B. 82 ride’ all point to *B, which usually corresponds to MC Departing tone (C). However, the MC form has the Level tone instead. Another case of tonal disagreement is ‘right’. Forms in Tai dialects all show *A, but the reconstructed MC form has the Departing tone. Note that the onset of this etymon is aberrant in CT varieties, i.e. /ɬa:A1/ instead of the expected /kʰwa:A1/. The CT forms may have been contaminated by *za:jC ‘left’, cf. Lungchow /ɬa:jC2/ (Gedney 1989a: 245). The form ‘to dig’ deserves special attention. It is very likely that the modern Tai forms have different etymological sources. Siamese as well as other SWT varieties go back to original voiceless onsets, pointing to 堀 kū from MC kʰwət. In contrast, the Po-ai and other NT varieties, the etymological source for ‘to dig’ seems to be 掘 jué from MC gjwət. As for CT, some varieties side with SWT while others side with NT. For example, the Lungchow form /kutDS2/ points to an earlier voiced onset but Yanshan Nung /kʰutDS1/ indicates an earlier voiceless onset. One very important generalization regarding the etyma with clear Chinese etymologies is that their MC forms all have voiced onsets. In addition to the etyma found in Table 3-6, a number other forms may have also come from Chinese. Table 3-7 presents voicing alternation items with possible Chinese origins. 83 Table 3-7 Voicing alternation items with possible Chinese origins Gloss ‘person’ Li Siamese Lungchow Po-ai *pʰ- pʰu:C1 pu:C1 pu:C2 ‘male’ ‘to come into contact’ ‘to braid, harrow’ ‘to carry, to hold’27 ‘hole, pit’ *pʰ- pʰu:C1 *tʰ- tʰu:kDL1 *f- fɯəA1 *tʰ- tʰɯ:A1 *kʰ kʰumA1 pʰu:C1 tʰukDL1 pʰɯ:A1 tʰɯ:A1 kʰumA1 pu:C2 tɯkDL2 fɯ:A2 tɯA2 kumA2 MC 父 bjuB or 夫 pju 夫 pju 觸 tśʰjwok 杷 ba(C) 戴 tậiC 坎 kʰậmB The etymon ‘person’ may be related to Chinese 父 fù from MC bjuC (Thurgood 2007: 251). According to Schuessler (2007: 243), 父 was used to mark respected persons engaging in certain activities, e.g. 農父 nóngfù ‘minister of agriculture’. Phonologically, this Chinese etymon is a good match for Tai ‘person’ but the meaning does not match very well. Another possibility is suggested by Qin (1992), who argues that this etymon comes from 夫 fū, which functions as a suffix for men of various occupations, e.g. 農夫 nóngfū ‘farmer’. This etymology offers a better match in meaning but has the Level tone instead of the expected Rising tone. Similarly, the Tai etyma for ‘male’ most likely also goes back to 夫 fū but suffers the same sound correspondence problem as ‘person’. Tai ‘to hit, to touch’ may go back to Chinese 觸 chù but the onset is unexpected, though note that the reconstructed OC form has *tʰ-. The etymon for ‘to braid’ may go back to Chinese 杷 pá but both the onsets and vowels are unexpected. Lastly, the etyma ‘to carry, to hold’, and ‘hole, pit’ may come from Chinese 戴 dài, and 坎 kǎn respectively but both the onset and tone are problematic. 27 This is author’s etymology. Thurgood (2007) links this etymon to 帶 dài ‘belt, girdle’ (MC tâiC). 84 The distribution of these etyma also reveals their post-PT origin. Both Gedney (1989a) and Li (1977: 36-39) claim that that there is a correlation between the voicing alternation and the classification of Tai languages. In particular, reflexes in SWT and CT are thought to point to earlier voiceless onsets, while their NT counterparts indicate earlier voiced onsets. Luo (1996), however, shows that these generalizations do not always hold because there are many cases in which some non-NT dialects also show reflexes of earlier voiced onsets. Following Luo, I show in Table 3-8 that different etyma show different distribution patterns with regards to the original voicing of the onsets. Table 3-8 Distribution patterns regarding original voicing of some voicingalternation items ‘to carry’ ‘to ride’ ‘bean’ ‘thick, dense’ SWT Siamese White Tai tʰɯ:A1 kʰi:B1 tʰuəB1 tʰi:B1 tɯ:A2 kʰwi:B1 tʰoB1 tʰiB1 CT Debao Lungchow NT Po-ai Wuming kʰwɤjB1 tʰuB1 tɤjB2 tʰɯ:A1 kʰwi:B1 tʰu:B1 tʰi:B1 tɯ:A2 kɯ:jB2 tu:B2 ti:B2 taɰA2 kɯjB1 tuB2 tojB2 The reflexes of ‘to carry’ in White Tai as well as many other SWT dialects unexpectedly shows /t-A2/ pointing to an earlier voiced onset, agreeing with NT rather than other SWT/CT dialects. For ‘to ride’, the Wuming form has /k-B1/ indicating an earlier voiceless stop, disagreeing with other NT varieties. For ‘bean’, SWT/CT varieties have /tʰ-B1/ from earlier voiceless onsets while NT dialects show /t-B2/ indicating an earlier voiced consonant. This is the expected pattern. Lastly, the modern 85 form for ‘thick, dense’ in Debao has /t-B2/ indicating an earlier voiced stop, siding with NT rather than SWT/CT. This kind of overlapping distribution is a signature of diffusion, and demonstrates that many etyma showing voicing alternation should be considered post-PT borrowings that were propagated throughout the Tai-speaking area after PT broke up into different languages. It has become clear that many voicing alternation items are post-PT borrowings as argued by Thurgood (2002a; 2007). It is now possible to consider how the Chinese etyma come to be reflected as voiced in some dialects but as aspirated in others. Taking the etymon ‘bean’ as a model case, I speculate that these Chinese etyma were borrowed into Tai when Chinese still preserved *NC̥ - clusters. Sagart (1999: 185-187) proposes OC *N-tho[k]-s for 荳 dòu < MC *dəuB. In this scenario, CT/SWT dialects simply dropped the *N- and borrowed this etymon as *tʰuəB1 while NT dialects changed the prenasalized stop into a plain stop, and borrowed this word as *duəB1. The borrowing may have happened after the OC period but definitely before the MC period because this *N-th- had become d- by the time of MC. If this scenario for ‘bean’ is correct, we can assume that other etyma in this set followed the same path. However, there are many cases of voicing alternation that cannot be treated as post-PT loans. Firstly, they do not show other irregularities beside the voicing alternation. Secondly, they are also found in other branches of Kra-Dai. These etyma thus may have been part of the PT lexicon. Table 3-9 shows etyma that have ProtoKam-Sui (Thurgood 1988), Proto-Hlai (Norquest 2007), or Proto-Kra (Ostapirat 2000) counterparts. 86 Table 3-9 Voicing alternation items that may have been part of PT their KraDai cognates ‘bitter’ ‘excrement’ ‘to arrive’ ‘ear’ Siamese kʰomA1 kʰi:A1 tʰɯŋA1 hu:A1 Po-ai hamC2 haiC2 taŋA2 rɯəA2 Kam-Sui *kam1 *ke4 *taŋ1 *kʰra:1 Hlai *C.ɦɤ:m *ɦa:jʔ Kra *kəmA *kajC *k.raA These four etyma seem to be good candidates for true Kra-Dai cognates. In Proto-Kam-Sui, they had voiceless stop onsets, except for *ke4 ‘excrement’. Similarly, the three Proto-Kra forms all had voiceless stops onsets. In contrast, the two ProtoHlai forms cited both show voiced fricatives. This situation is in sharp contrast with those that can be shown to be Chinese loans, which had voiced onsets in MC. Therefore, it is unclear whether the voiceless reflexes in Siamese and other non-NT languages, or the voiced reflexes in Po-ai and the rest of NT preserved the original voicing. In fact, it is possible that the original onsets were complex and consisted of both a voiced and a voiceless consonant. Similar to Ferlus (1990), I tentatively posit sesquisyllabic clusters of the type *C̬ .C̥ - for voicing alternation items that cannot be explained as borrowings. In this scenario, ‘bitter’, ‘excrement’, and ‘to arrive’ should be reconstructed at the PT level as *C̬ .qɤmA, *C̬ .qɯjC and *C̬ .tɤŋA respectively (see §4.4.3). One exception is ‘ear’, for which I reconstruct *k.rwɯ:A and speculate that the *k.- was a prefix that was lost early on in NT but preserved in CT and SWT. Note that ‘to arrive’ might ultimately be related to PAN *datəŋ ‘to arrive’. In this section, I have shown that the so-called voicing alternation is not a unified phenomenon. On the contrary, it consists of two distinct sets of data. The first set consists of those that are post-PT borrowings and must not be reconstructed for PT. 87 Not surprisingly, most of the etyma came from Chinese. The second set consists of etyma that cannot be explained away as loans and may have been part of the PT lexicon. I tentatively account for these cases by positing sesquisyllabic clusters as onsets of these etyma. 3.5 Lack of aspiration in PT Aspiration has long been viewed as contrastive in PT (Haudricourt 1948; Li 1977; Nishida 1954, 1975a). In this conventional view, SWT and CT languages preserve the aspiration intact while their NT relatives lost it altogether. At first glance, this view of the PT consonant system seems to account for the dialectal diversity straightforwardly. However, Haudricourt (1963b: 157) and Ferlus (1990)28 believe that aspiration secondarily developed in SWT and CT varieties out of original consonant clusters. A close examination of the data reveals that viewing contrastive aspiration as a characteristic of Proto-Tai is problematic. The first problem is that loss of aspiration is not a common sound change. Recognizing that cases of de-aspiration are extremely rare, Ferlus (1990) argues that loss of aspiration is particularly implausible in Northern Tai languages, which have been heavily influenced by Chinese, where aspiration is very robust. The second problem is that the evidence for Li’s simple aspirated stops *pʰ-, *tʰ-, *cʰ-, and *kʰ- is not as robust as it appears to be. Many of the etyma listed under these onsets show the so-called voicing alternation, and must be considered either later loans or reflexes of complex onsets (see §3.4). Moreover, etyma on Li’s lists that invariably have aspirated onsets are attested by only a few examples, most of which are questionable or irregular (Gedney 1989a). Comparison with Kam-Sui also further casts doubt on the view that 28 In a personal communication, Ferlus explains that aspiration in his reconstruction of PT is not contrastive but serves as a “junction” between the stop and the medial *-r-, i.e. *pʰr- is in fact *pr-. 88 PT aspirated consonants were contrastive. Li (1965: 55) noted that etyma with aspirated onsets in Kam-Sui do not often agree with the Tai aspirated onsets. More radically, Thurgood (1988) claims that aspiration in Kam-Sui and other branches is not an inherited trait from Proto-Kra-Dai. Similarly, Liang and Zhang (1993; 1996) claim that aspiration developed independently in different Tai varieties from various complex onsets, e.g. *pr-, *tl-, *tr-, *tr-, *pw-, *xp-, *xpl-, and *xt-. In this section, I argue that PT lacked contrastive aspiration and that aspirated stops found widely among modern Tai languages are post-PT innovations. Specifically, I first show that etyma conventionally reconstructed with simple aspirated onsets that do not go back to original clusters were not part of the PT lexicon. I then propose that aspiration in CT and SWT developed mainly from aspiration of clusters with medial *-r-. 3.5.1 Li’s simple aspirates as post-PT innovations Although Li (1977) lists a number of items under simple aspirated stops *pʰ-, *tʰ-, *cʰ-, and *kʰ-, reconstructing contrastive aspiration for PT is in fact very problematic. As discussed earlier, the majority of these etyma show voicing alternation and therefore clearly do not go back to PT simple aspirates. Moreover, a number of etyma that Li reconstructed with *kʰ- should in fact be reconstructed with uvular onsets (see §3.3). In this subsection, I show that most of the remaining items must be viewed as post-PT lexical innovations, either borrowings or forms derived after the establishment of the contrastive aspiration. These etyma can be divided into two groups—those that are found only in SWT and CT and those that are found in all the three groups. Table 3-10 lists etyma with aspirated onsets that are not found in NT. 89 Table 3-10 Forms with Li’s simple aspirated onsets not found in NT Gloss ‘carpenter's square’ ‘sing’ ‘cloudy, dark’ ‘turban’ ‘to dry on fire’ ‘frog’ ‘to coil’ ‘mother's younger brother’ ‘spicy hot’ ‘not exact, wrong’ ‘to sweep’29 ‘to knit’30 Li *cʰ*kʰ*kʰ- *kʰ*kʰ*kʰ*kʰ*kʰ- *pʰ*pʰ*pʰ*tʰ- Siamese cʰa:kDL1 kʰapDS1 kʰa:ŋA1 kʰiətDL1 *kʰotDS1 Lungchow kʰamA1 kʰɯnA1 kʰa:ŋB1 kʰi:tDL1 kʰu:C1 Po-ai pʰetDS1 pʰitDS1 pʰɛ:wC1 tʰakDS1 pʰitDS1 The etymon ‘carpenter’s square’ is most likely a Siamese innovation. It is found only in Siamese and Lao. The Dioi form cited by Li (1977: 167) is probably not related. It may be related to Chinese 尺 chǐ (< MC tśʰjäk) ‘foot measure’. The forms ‘to sing’, ‘turban’, and ‘to coil’ are reflected with /x-/ in White Thai, pointing to either earlier *x-, *q-, or *χ-, rather than *kʰ-. The /h-/ in the Saek reflex for ‘to sing’ also supports this interpretation. As for ‘turban’ and ‘mother’s younger brother’, it is difficult to identify their original onsets because they are found exclusively in CT dialects, which have all lost the distinction. However, these etyma are ultimately from 29 For Po-ai, Li (Li 1977: 64) gives /pewC1/ ‘to move residence’. In addition to the semantic problem, the short vowel is also unexpected. Therefore, this form is not related. 30For NT, Li (Li 1977: 102-103) cites Po-ai /takDS1/ ‘knot’, which may not be related. 90 Chinese 巾 jīn (< MC kjen) and 舅 jiù (< MC gjəuB) respectively. Note that ‘turban’ is mostly likely borrowed via Vietnamese, cf. xăn. The entry ‘to dry on fire’, found only in CT and SWT only, shows irregularity with respect to the tone. While the Siamese form goes back to *A, the Lungchow form indicates *B. The NT forms for ‘frog’ go back to PT *krwe: but are not related to their CT/SWT counterparts. The form ‘to knit’ might not be found outside of Siamese. Manamaivibool (1975: 346) links this word to Chinese 織 zhī (< MC tśjək) ‘to weave’. The forms for ‘spicy hot’ are found only in SWT and CT, assuming that the Saek forms for these etyma are loans from Lao or Thai. Similarly, ‘not exact’ is exclusively a SWT word. While ‘spicy hot’ might have been borrowed from an Austronesian source, cf. Malay /pahit/ ‘bitter’, the etymology of ‘not exact’ is unclear. The crucial point is that the etyma in this set only occur in CT and SWT, never in NT. Many of them point to initials other than *kʰ-, i.e. *x-, *q-, or *χ-. Although it is not possible to explain the etymological sources of every etymon, the words in this set cannot be regarded as cases of retention from PT on the part of CT and SWT. This is because they are the only instances of Li’s simple aspirates. If these particular forms were retentions, we also would expect to see in NT a sizeable number of other etyma shared with either CT or SWT that fit this correspondence pattern. Since no such item is found, these etyma must be considered post-PT vocabulary that were innovated after CT and SWT dialects had developed contrastive aspiration through independent processes (see §3.3 and §3.5). Unlike the set of data just discussed, the second set, given in Table 3-11 consists of etyma reconstructed with simple aspirated stops by Li that are shared by SWT, CT, and NT. 91 Table 3-11 Forms with Li’s simple aspirated onsets that are found in all the three groups Gloss ‘to split’ ‘to roll, to twist’ ‘clf. for boards, flat objects, etc.’ ‘charcoal’ ‘to pull out’ ‘to retreat’ ‘to remove’ ‘granary’ ‘to tear’31 ‘to lie athwart, broad’32 Li Siamese Lungchow Po-ai *pʰ- pʰa:B1 pʰa:B1 paB1 *pʰ- pʰanA1 pʰanA1 panA1 *pʰ- pʰɛnB1 pʰe:nB1 pe:nB1 MC 破 pʰwâC 翻 pʰjwɐn 片 pʰienC *tʰ- tʰa:nB1 *tʰ- tʰɔ:nA1 *tʰ- tʰɔ:jA1 *tʰ- tʰɔ:tDL1 *cʰ- cʰa:ŋA1 *cʰ- cʰi:kDL1 *kʰw- kʰwa:ŋA1 tʰa:nB1 tʰo:jB1 tʰutDS1 ɕa:ŋA1 ɕi:kDL1 kʰwa:ŋB1 ta:nB1 tʰo:nB1 tʰo:jB1 to:tDL1 ɕa:ŋA1 ɬi:kDL1 kwa:ŋB1 炭 tʰânC 退 tʰwậiC 脫 tʰwât 倉 tsʰâŋ 拆 ṭʰɐk 橫 ɣwaɨjŋ Although the etyma in Table 3-11 are found in all three groups and are reflected with the expected onsets, they cannot be reconstructed at the PT level. First of all, a few of them show unexpected vowels or tones. For ‘to pull out’, the tones in Siamese and Po-ai do not agree. Similarly for ‘to retreat’ and ‘to lie athwart, broad’, Siamese points to *A while Lungchow and Po-ai indicate *B. As for ‘to remove’, the Lungchow form has short /u/ instead of the expected long /o:/. More crucially, it cannot be coincidental that all but one has clear Chinese etymologies. This strongly argues for treating these etyma as post-PT borrowings. In other words, they were not part of PT but were borrowed from Chinese separately after the break-up of PT into 31 This etymology is my own. Pulleyblank reconstruct tʰrɛ:jk. 32 This etymon is not listed in Li (1977) but follows the same pattern. 92 different branches. In CT and SWT, the borrowing must have occurred after contrastive aspiration had emerged (see §3.5.2 below). This allows these Chinese loans to maintain their aspiration when incorporated into the recipient Tai dialects. On the other hand, NT dialects, which never acquired contrastive aspiration, lost the Chinese aspiration in the process of loanword adaptation. 3.5.2 Aspiration of original PT clusters Having established in the previous section that Li’s simple aspirated onsets in fact cannot be reconstructed for PT, we must then expect clusters with aspirated stops found in modern Tai languages to have developed secondarily. In the conventional reconstruction of PT onsets (Haudricourt 1948; Li 1977; Sarawit 1973), the aspirated stops can be combined with medial *-r-, *-l-, and *-w- to form complex onsets. In contrast to simple aspirated onsets discussed above, evidence for clusters consisting of an aspirated consonant followed by a medial *-r- are very robust. This is suspicious because clusters should consist of simple consonants that exist independently in the language. This indicates that Li’s clusters with aspirated stops plus medial *-r- must in fact be reconstructed differently. Similarly, Ferlus (1990) does not reconstruct simple aspirates but posits clusters of type aspirated stops plus *-r-. However, he explains in a personal communication that he considers the aspiration in those clusters to be a “junction sound”. In this section, I argue that native Tai etyma that are reflected with aspirated onsets in CT and SWT languages must have had medial *-r- at an earlier stage. The sets of data in Table 3-12 illustrate this process. It is this medial *-r- that caused PT plain voiceless stops to become aspirated. From the point of view of this proposal, Ferlus’s system represents an intermediate stage when the aspiration induced by the medial *-r- is still sub-phonemic. 93 Table 3-12 PT clusters with medial */-r-/ Gloss ‘vegetable’ ‘forehead’ ‘head hair’ ‘headlouse’ ‘tail’ ‘stone’ ‘plough’ ‘to wait’ ‘to seek’ ‘six’ ‘to laugh’ Li *pʰr*pʰr*pʰr*tʰr*tʰr*tʰr*tʰl*tʰl*xr*xr*xr- Siamese pʰakDS1 pʰa:kDL1 pʰomA hawA ha:ŋA hinA tʰajA tʰa:C ha:A hokDS1 huəA Debao pʰjakDS1 pʰja:kDL1 pʰjamA tʰawA tʰa:ŋA tʰɤnA tʰajA tʰa:C kʰja: kʰjokDS1 kʰu:A Po-ai pjakDS1 pja:kDL1 pjɔmA lawA lɯ:ŋA lamA sajA sa:C la:A lɔkDS1 li:wA Saek pʰrakDS1 pʰra:kDL1 pʰram A rawA ruəŋA tʰajA tʰa:C ra:A rɔkDS1 ruəwA PT *pr*pr*pr*tr*tr*tr*cr*cr*kr*kr*kr- In the table above, Siamese represents SWT, Debao represents CT, and Po-ai and Saek represent NT. The first set of etyma must go back to PT *pr- rather than *pʰl/r- as reconstructed by Li. The medial *-r- had an effect on the preceding *p- such that it became aspirated in Siamese, Debao, and Saek. This process of aspiration apparently did not occur in Po-ai. For the second set, the PT cluster *tr-, which correspond to *tʰr- in Li’s system became aspirated in Debao before the loss of *-r-. In Siamese and Saek, the *tr- may have first also became aspirated to *tʰr- and then developed into a voiceless *ʰr-. In contrast, it is clear that in Po-ai the *tr- did not aspirate but became simplified directly to *ʰr-, which is now reflected as /l-/. In other words, the initial *t- was lost but the voicelessness is preserved on the medial *-r-. This is because, as evidenced in the case of PT *pr-, Po-ai did not participate in the 94 aspiration process. Skipping to the fourth set, for which Li reconstructs *xr-, we see a very similar situation. In all languages shown except for Debao the velar onset in the cluster *kr- was lost, resulting in a voiceless sonorant *ʰr-. Lastly, the third set presents cases of PT *cr- for Li’s *tʰl-. The cluster simply lost the medial *-r- in Po-ai but became *cʰr- through the aspiration process in Siamese, Debao, and Saek. This palatal cluster later became *tʰr-, and then *tʰ-. Another source of aspiration in CT dialects is sesquisyllabic clusters that at some stage simplified to tautosyllabic cluster with medial *-r- (see Chapter 4). In Li’s system, most of these clusters are reconstructed as plain stops plus medial liquids; however this reconstruction cannot account for the observed data. For example, Li’s *tr- cannot account for the labiality of the Saek forms, which should in fact be reconstructed with a labial consonant. Following Haudricourt (1956), and Ferlus (1990), I reconstruct this PT complex onset as *p.t-, cf. Lakkja *plejA1 ‘to die’ and PAN *matáy ‘to die’. In Debao, this *p.t- must have first become *tr- due to monosyllabicizing pressure and assimilation of the initial stop to *-r-, merging with original *tr-. In Saek, *p.t- became *pr- rather than *tr- but was kept apart from original PT *pr- which had already gone through the process of aspiration. Table 3-13 gives examples of sesquisyllabic clusters that are reflected as aspirated clusters in modern languages. 95 Table 3-13 Examples of PT sesquisyllabic clusters that are reflected with aspiration Gloss ‘to die’ ‘eye’ ‘grasshopper’ ‘centipede’ ‘to imprison’ ‘to drive away’ Li Siamese Debao *tr- ta:jA tʰa:jA *tr- ta:A tʰa:A *tr- takDS2 -t tʰakDS1 *kʰr- kʰepDS1 *kʰl-A kʰaŋA1 kʰjaŋA1 *kʰl-A kʰapA1 Po-ai ta:jA ta:A takDS1 ɬipDS1 caŋA1 capA1 Saek pra:jA pra:A tʰripDS1 tʰraŋA1 PT *p.t*p.t*p.t*q.s*k.r*k.r- Similar to the case of *p.t-, PT *q.s- in ‘centipede’ became *q.r- merging with original *q.r-, which is now reflected as /kʰ-/ in Siamese and /kʰj-/ in Debao. In Po-ai, *q.s- simplified to *s-, which is now reflected as /ɬ-/. In addition, some sesquisyllabic clusters that had *.r- as onset of the major syllable were reduced to tautosyllabic clusters with medial *-r- as illustrated by the case of ‘to imprison’ and ‘to drive away’ above. In these cases, the PT sesquisyllabic cluster *k.r- was reduced to a tautosyllabic cluster *kr- before becoming *kʰr-. This intermediate cluster is now reflected as /kʰ-/ in Siamese, /kʰj-/ in Debao, /c-/ in Po-ai, and /tʰr-/ in Saek33. For more discussion on reduction of sesquisyllabic clusters, refer to §4.5. Further evidence for this proposal comes from Kam-Sui languages. Etyma that are reconstructed with *-r- clearly correspond to unaspirated clusters in Thurgood (1988)’s preliminary reconstruction of Proto-Kam-Sui. For example, PT *prɤmA ‘head hair’, *pra: ‘stone mountain’, *cra:C ‘to wait’, and *krɯəw ‘to laugh’ correspond to *pram1, *pra1, *kra3, and *kru1 in Proto-Kam-Sui respectively. These Kam-Sui forms 33 Saek changed *kʰr- to /tʰr-/ regularly. 96 strongly support the view that the aspiration in clusters with medial *-r- developed secondarily in the CT and SWT of daughters of PT. The lack of aspiration in the NT reflexes of these etyma is therefore an inherited trait from Proto-Kra-Dai. Aspiration induced by medial -r- is a not an unusual sound change. Many cases have been reported, especially in SEA. In Tai Yuan34, the SWT dialect of Thailand, clusters consisting of an unaspirated stop followed by a medial *-r- 35 are now pronounced as aspirated stops, e.g. /pʰe:tDL1/ from earlier *pre:tDL (< Sanskrit preta-) ‘hungry ghost, /kʰiŋB1/ ‘to be suspicious’ from *kriŋB, /kʰa:pDL1/ ‘to pay respect’ from *kra:pDL, and /pʰɛ:A1/ ‘to vary’ from *prɛ:A. These words are still pronounced as /pre:tDL1/, /kriŋB1/, /kra:pDL1/, and /prɛ:A1/ in Siamese. Moreover, the first three etyma are spelled as pre:t, kriŋ, and kra:p in Tai Yuan palm-leaf manuscripts respectively. Another example of aspiration induced by medial *-r- is from Khmer. In the colloquial Phnom Penh dialect, initial stops /p-, t-, k-/ when followed by /-r-/ are pronounced as [pʰ-, tʰ-, kʰ-], e.g. /pram/ ‘five’ pronounced as [pʰeam], and /prap/ ‘to tell’ as [pʰeap] (Guion & Wayland 2004; Wayland & Guion 2005). In other words, the /-r-/ aspirated the initial stops before dropping off, causing a change in the vowel quality. Phonetically, aspiration of *-r- clusters is a very plausible sound change. Wayland and Guion (Guion & Wayland 2004; Wayland & Guion 2005) propose with respect to the Khmer case just discussed that aspiration in Cr- clusters helps facilitate the production of the medial /-r-/. From an acoustic point of view, a period of aspiration develops between the stop and the medial /-r-/ because the glottis remains open throughout the production of the cluster. Moreover, from an aerodymamic point of view, the aspiration increases the airflow needed in the production of the medial /-r-/. Specifically, spread glottis, which is associated with aspiration, allows for relatively 34 The modern Tai Yuan forms are taken from Wichiankhiaw et al (1996). 35 These clusters are found in loanwords only. Original PT clusters of this type had gone through several changes by the time of PSWT. 97 strong and high volume of airflow needed to initiate the trill. Therefore, the aspiration of original Cr- clusters found in CT and SWT dialects is a natural sound change. The view that PT had no distinctive series of aspirated onsets has a few structural advantages over the conventional view that four contrastive phonation types existed at the PT level. Firstly, it explains curious gaps in the PSWT stop system I have discussed elsewhere that would otherwise be unexplained (Pittayaporn 2008b; to appear-b). Despite the rich PSWT consonant inventory, only labial and velar stops can be combined with medial *-l- and *-r- as illustrated in Table 3-14. The unaspirated stops were combined only with medial *-l- and never with *-r-. In contrast, aspirated stops only took medial *-r-, as the only aspirated cluster in PSWT is *kʰr-. Table 3-14 Clusters consisting of a stop plus a liquid in PSWT labial velar *Cʰ- *kʰr*C- *-r*C̬ - *br- *gr*Cʰ*C- *-l- *pl- *kl*C̬ - *bl- *gl- In the proposed reconstruction, these gaps are accounted for straightforwardly as consequences of the aspiration of PT *pr-, and *kr-. Specifically, these two clusters aspirated and became *pʰr-, and *kʰr-, each leaving a gap behind. These intermediate clusters then simplified to *pʰ- and *ʰr- in PSWT respectively. The gap in the velar series left behind by the simplification of *kʰr- was filled by orginal *qr-, *k.r-, and *q.s all of which became *kʰr- in PSWT (see Chapter 4). Similarly, the fact that 98 aspirated clusters with medial *-l- did not exist in PSWT is accounted for very nicely as in this proposal PT *-l- did not aspirate its preceding stops. Note that, on this view, the PSWT simple aspirated series came from different sources, including PT *cr-, PT *qr-, and later Chinese loans, etc. This scenario is schematized in Table 3-15. Table 3-15 Schematization of the development of PSWT stop system from PT *C*Cʰ*C̬ *C*Cʰ- -l*C̬ - *C*Cʰ- -r*C̬ - labial pʰ*p*b*pl- *bl(PT *pr-) *br- velar *kʰ*k*g*kl- *gl*kʰr(PT *kr-) *gr- loanwords, voicing alternation loanwords PT *qr-, *q.r-, *ksPSWT *ʰr- The second advantage of the reconstruction proposed above is that the PT palatal series is no longer defective when compared to other places of articulation. Each of the five places of articulation in PT had a plain voiceless stop, and a voiced stop, except for the labial and alveolar series which also had an implosive. All these PT stops can be combined with either an *-l- or an *-r-. In this system, the fact that *cʰ- is not reconstructible for PSWT is then explained as a result of the change from PT *cr- to *tʰ- instead of *cʰ-. This change can be understood as a ban against palatal clusters, which are phonologically marked. 99 Another appeal of the proposed PT reconstruction is that it offers an explanation for why some Chinese loanwords with aspirated initials unequivocally go back to earlier unaspirated stops in Tai. Among the Sino-Siamese vocabulary identified by Manomaivibool (1975) a few etyma have aspirated initials in Chinese but unaspirated initials in Siamese as shown in Table 3-16. Table 3-16 Forms showing aspirated initials in Chinese but unaspirated in Tai. Gloss ‘seven’ ‘lung’ ‘duck’ Siamese cetD1 pɔ:tD1 pet D1 Lungchow citD1 pɯtD1 Po-ai ɕɛtD1 pɯtD1 pɯtD1 pɯtD1 MC tsʰjət pʰjwɐiC pʰjiet OC 七 *tshit 肺 *phots 鴄 *phit Manomaivibool (1975: 124) notes that such alternations in the correspondences between Chinese and Thai forms occur frequently and hints that it may possibly indicate a morphological process yet to be identified. In contrast, the reconstruction proposed here explains this phenomenon quite nicely by treating these etyma as early loans into PT before aspiration was developed. Because PT did not have aspirated stops, Chinese words with aspirated stops must have gone through a process of loanword adaptation whereby the initials were converted into the most similar unaspirated stop. The implication is then that all Chinese borrowings in Tai languages that show aspiration were incorporated after the break-up of PT. In sum, we can see that all of Li’s clusters with aspiration must be given an alternative set of reconstructions. For example, *pʰr- and *tʰr- must be replaced by *pr- and *tr- while *tʰl- must be rewritten as *cr-. Li’s clusters that consist of a plain stop followed by medial *-r- must be reconstructed as sesquisyllabic onsets. For instance, Li’s *tr- corresponds to *p.t- in the current system while some of the etyma 100 with *kr- must in fact be reconstructed with *q.s-. Some clusters with fricatives must also be replaced by complex onsets with plain stops followed by *-r-. For example, Li’s *xr- is reconstructed as *kr- in the system proposed here. 3.6 Reconstructions of PT simple initials Etyma from Siamese (SI), Lungchow (LC), and Yay (Y), are provided in each section to represent the three clusters of the modern varieties. Remember that this is not to say that they represent three genealogical subgroups of Tai (See §1.4.2.3). 3.6.1 Voiceless stops There were altogether five voiceless stops that could function as onsets in PT: *p-, *t-, *c-, *k-, and *q-. Except for the palatal *c- and the uvular *q-, these PT stops are kept intact in all modern varieties. The glottal stop is *ʔ- is grouped with the glottalized/implosive consonants because it always pattern with them with respect to tonal splits and mergers. The etyma that go back to these PT voiceless stops all have tones in the first series. 3.6.1.1 *pPT voiceless labial stops are preserved as /p-/ without change in all attested modern Tai varieties. Etyma that go back to this PT initial show tones in the first series, which indicate that this PT phoneme was voiceless. PT *p- in the current reconstruction corresponds to Li’s *p- (1977: 61-63). Examples of PT etyma with *pare given in Table 3-17. 101 Table 3-17 Etyma with PT *p- Gloss ‘parent’s older sister’ ‘mouth’ ‘year’ ‘eight’ ‘to peel’ PT *pa:C *pa:kD *pi:A *pe:tD *po:kD SI pa:C1 pa:kDL1 pi:A1 pɛ:tDL1 pɔ:kDL1 LC pa:C1 pa:kDL1 pi:A1 pe:tDL1 po:kDL1 Y pa:C1 pa:kDL1 pi:A1 petDL1 pokDL1 Others 3.6.1.2 *tPT voiceless alveolar stops are reflected as /t-/ in all attested modern Tai varieties. Etyma that go back to this PT initial show tones in the first series, which indicate that this PT phoneme was voiceless. It is also reconstructed as *t- in Li (1977: 97-102). Examples of PT etyma with *t- are given in Table 3-18. Table 3-18 Etyma with PT *t- Gloss ‘to scoop’ ‘low’ ‘door’ ‘to drop’ ‘to nibble’ PT *takD *tamB *tu:A *tokD *to:tD SI takDS1 tamB1 tu:A1 tokDS1 tɔ:tDL1 LC takDS1 tamB1 tu:A1 tukDS1 to:tDL1 Y takDS1 tamB1 tuA1 tokDS1 totDL1 Others 3.6.1.3 *cThe reflexes of the PT voiceless palatal stop in modern varieties differ considerably. In most SWT and CT dialects and Saek, it is preserved as a palatal stop 102 /c-/, which is usually phonetically affricated [tɕ-]. Some examples of dialects that have this conservative reflex are Siamese, Lao, Black Tai, Aiton, Lungchow, Lungming, Western Nung, etc. In many dialects, however, it has become a true affricate /ts-/. Some modern dialects that show this reflex include Tai Ya, Debao, Shangsi, Chongzuo, Qinzhou, Fusui, Wenma, Liujiang, Rong’an, etc. Many Shan dialects as well as many NT varieties have gone one step further and turn the affricate into a fricative, either /s-/ or /ɕ-/. Southern Shan, Tai Mao, Po-ai, Yay, Longsheng, and Hechi are examples of varieties that show /s-/, while Wuming, Tiandong, Lingyue, and Du’an represent dialects that show /ɕ-/. Li (1977: 164-167) reconstructs an affricate *tɕ-36 (see §3.2). Examples of PT etyma with *c- are given in Table 3-19. Table 3-19 Etyma with PT *c- Gloss ‘breath, heart’ ‘to kiss’ ‘to sink’ ‘insipid’ ‘seven’ PT *caɰA *cupD *cɤmA *cɯ:tD *cetD SI cajA1 cu:pDL1 comA1 cɯ:tDL1 cetDS1 LC caɰA1 cupDS1 citDS1 Y sɯA1 sam satDS1 Others Rong’an /tsupDS1/ Rong’an /tsamA1/ Rong’an /tsitDS1/ Rong’an /tsatDS1/ 3.6.1.4 *kPT voiceless velar stops are reflected as /k-/ in all attested modern Tai varieties. Etyma that go back to this PT initial show tones in the first series, which indicate that this PT phoneme was voiceless. It is also reconstructed as *k- in Li (1977: 186-192). Examples of PT etyma with *k- are given in Table 3-20. 36 *č- in Li’s notation 103 Table 3-20 Etyma with PT *k- Gloss ‘crow’ ‘to take a bite’ ‘old (for living being)’ ‘frog’ ‘to hug’ PT *ka:A *katD *ke:B *kɤpD *ko:tD SI ka:A1 katDS1 kɛ:B1 kopDS1 kɔ:tDL2 LC ka:A1 katDS1 ke:B1 kupDS1 ko:tDL2 Y katDS1 ceB1 kapDS1 kotDL2 Others Wuming /ka:A1/ For ‘crow’, Yay and most other NT dialects has /ʔaA1/. This form is probably not related to PT *ka:A, but most likely a case of onomatopoeia. Yay regularly palatalized velar consonants preceding front vowels37, thus showing /ceB1/ for ‘old (for people)’. 3.6.1.5 *qThe PT voiceless uvular stop is reflected either as /kʰ-/, /x-/, /h-/, or /k-/ in different varieties. In CT, it is generally reflected as /kʰ-/ as in Lungchow, Lungming, Debao, Chongzuo, Daxin, etc. In a few CT dialects including Shangsi, Yongnan, Fusui, Qinzhou, and Long’an, as well as all NT dialects, i.e. Yay, Wuming, Bouyei, Longsheng, etc., however, this PT onset merged with *k- to give modern /k-/. Note that *q- merged into *k- in Yay and became palatalized in front of front vowels. PT *q- remains *q- in PSWT but later became affricated to *χ- (Pittayaporn 2008b; to appear-b). This *χ- is now reflected as /kʰ-/ in Siamese, Lao, and Black Tai, and /x-/ in some Shan dialects such as Dehong. Only in a handful of SWT dialects is *q- kept 37 PT uvular consonants merged with the velar consonants in time to participate in this change (see §3.6.1.5). 104 distinct from *kʰ-, which is the PSWT reflex of PT *C̬ .k- and *C̬ .q-. For example, in White Tai, PSWT *q- is reflected as /x-/ and PSWT *kʰ- as /kʰ-/. In contrast, the Kapong dialect of Phu Thai shows /h-/ for *q- but /kʰ-/ for PSWT *kʰ- (< PT *C̬ .kand *C̬ .q-). Refer to §3.3 and §3.4 for discussion. Table 3-21 presents examples of etyma reconstructed with PT *q-. Etyma with PT *q- are reconstructed with *kʰ- or *x- in Li (1977: 192-197, 208-214). Table 3-21 Etyma with PT *q- Gloss ‘to kill’ ‘torn’ ‘horn’ ‘needle’ ‘arm’ PT *qa:C *qa:tD *qawA *qemA *qe:nA SI kʰa:C1 kʰa:tDL1 kʰawA1 kʰemA1 kʰɛ:nA1 LC kʰa:C1 kʰa:tDL1 kʰimA1 kʰe:nA1 Y ka:C1 ka:tDL1 kawA1 cimA1 cenA1 Others Kapong /ha:/ Kapong /ha:tDL1/ Kapong /hawA1/ Kapong /hɛ:nA1/ 3.6.2 Implosive/glottalized Two implosive stops are posited for PT—*ɓ-, and *ɗ-. These phonemes are interesting in terms of tonal development. They usually pattern with voiceless onsets in tonal splits but there are plenty of cases where they pattern with voiced onsets, especially for tone *A (see §6.2 and §6.4). In addition to these two implosives, two glottalized consonants *ʔ- and *ˀj- are reconstructed. These two consonants always pattern with the implosives with respect to tonal splits and mergers. Phonologically, the glottalized palatal glide *ˀj- can be considered a palatal counterpart of the implosives. 105 3.6.2.1 *ɓThe PT labial implosive *ɓ- remains an implosive or preglottalized stop in many dialects, such as Debao and Wuming. In most dialects, it is simplified to plain voiced /b-/. These dialects include Siamese, Lao, Black Tai, White Tai, Lungchow, Leiping, Saek, Yay, Qiubei, Huanjiang, Bouyei, etc. Some dialects have turned this sound into a sonorant. For example, Southern Shan, Lungming, Yongnan, Fusui, Yongbei, and Long’an all reflect PT *ɓ- as /m-/. Similarly, Longsheng’s reflex of PT *ɓ- is /w-/. Some dialects such as Yuanyang and Menglian have gone one step further and turned this resulting /w-/ into /v-/. That the implosive *ɓ- became sonorants in some languages is consistent with Clements and Osu (2002)’s view that implosives are neither obstruents nor sonorants. Li (1977: 68-71) reconstructs *ˀb- for this onset. Table 3-22 gives examples of PT etyma with *ɓ-. Table 3-22 Etyma with PT *ɓ- Gloss ‘village’ ‘unmarried man’ ‘leaf’ ‘to fly’ ‘disgusted’ PT *ɓa:nC *ɓa:wB *ɓaɰA *ɓinA *ɓɯəB1 SI ba:nC1 ba:wB1 bajA1 binA1 bɯəB1 LC ba:nC1 ba:wB1 baɰA1 binA1 bɯ:B1 Y ba:nC1 ba:wB1 baɰA1 binA1 bɯəB1 Others Wuming /ˀba:nC1/ Wuming /ˀba:wB1/ Wuming /ˀbaɰA1/ Wuming /ˀbinA1/ Wuming /ˀbɯB1/ Note a semantic change in Siamese from ‘village’ to ‘house’. In addition, the Siamese form for ‘unmarried man’ now means ‘servant’. It is also found in the compound /ca:wC1 ba:wB1/ ‘groom’. 106 3.6.2.2 *ɗThe PT alveolar implosive *ɗ- remains an implosive or preglottalized stop in many dialects, such as Debao and Wuming. In most dialects, it is simplified to plain voiced /d-/. These dialects include Siamese, Lao, Black Tai, White Tai, Lungchow, Leiping, Saek, Yay, Qiubei, Huanjiang, etc. Parallel to *ɓ-, some dialects have turned this sound into a sonorant. For example, Southern Shan, Tai Nüa, Liujiang, Yishan, Rong’an, and Bouyei all reflect PT *ɗ- as /l-/. In Qinzhou, Yongnan, and Long’an, the PT *ɗ- is reflected as /n-/. In addition, Longsheng is quite unique in having /r/ as reflex of PT *ɗ-. Like its labial counterpart, the fact that *ɗ- is reflected as a sonorant in some dialects but an obstruent in others is consistent with the view that implosives are nonobstruent and non-sonorant (Clements & Osu 2002). Li (1977: 107-111) reconstructs *ˀd- for this onset. Table 3-23 gives examples of PT etyma with *ɗ-. Table 3-23 Etyma with PT *ɗ- Gloss ‘to scold’ ‘nose’ ‘good’ ‘forest’ ‘late at night’ PT *ɗa:B *ɗaŋA *ɗɤjA *ɗoŋA *ɗɯkD SI da:B1 daŋC1 -t di:A1 -v doŋA1 dɯkDS1 LC da:B1 daŋA1 dajA1 doŋA1 Y da:B1 daŋA1 di:A1 doŋA1 dakDS1 Others Wuming /ˀda:B1/ Wuming /ˀdaŋA1/ Wuming /ˀdojA1/ Debao /ɗakDS1/ The meaning of ‘nose’ has been narrowed in Siamese to ‘nose bridge’, which also shows an unexpected tonal reflex, i.e. C1 rather than A1. 107 3.6.2.3 *ˀjThis glottalized glide *ˀj- is the palatal counterpart of the implosives *ɗ- and *ɓ-. It is usually simplified to a simple /j-/, as in Siamese, Lao, Black Tai, White Tai, Lungming, Lungchow, Shangsi, Long’an, Jingxi, Wenma, etc. In Qiubei, the /j-/ has been hardened to /dʑ-/. Moreover, younger speakers of Sapa and Bao Yen, both spoken in Vietnam, now have a fricative pronunciation /z-/. In some dialects, it is still preserved as /ˀj-/. These conservative dialects include Wuming, Daxin, Pingguo Tiandong, Donglan, etc. Li (1977: 181-185) also reconstructs *ˀj- for this set of data. Examples of PT *ˀj- are given in Table 3-24. Table 3-24 Etyma with PT *ˀj- Gloss ‘medicine’ ‘hungry’ ‘to be at’ ‘to stretch’ ‘to roast’ PT SI *ˀjɯa:A ja:A1 *ˀjɯəkD ja:kDL1 *ˀju:B ju:B1 *ˀjiətD jiətDL1 *ˀjɯa:ŋC ja:ŋC1 LC ja:A1 ja:DL1 ju:B1 ji:tDL1 Y jiəA1 jiəkDL1 ju:B1 jiəŋC1 Others Wuming /ˀjɯA1/ Wuming /ˀjɯəkDL1/ Wuming /ˀjawB1/ Wuming /ˀjiətDL1/ Wuming /ˀjɯəŋC1/ In addition to *ˀj-, Li (1977: 178-181) also reconstructs *j- for the following set of data. He includes more etyma than those listed in Table 3-25 but most of them show irregularities, only occur in SWT, or should be reconstructed with other onsets. 108 Table 3-25 Some etyma that Li reconstructs with *j- Gloss SI ‘paternal grandmother’ ja:B2 ‘to mend’ ja:A1 ‘difficult’ ja:kDL2 ‘web’ jajA2 ‘slack’ ja:nA2 LC ja:A2 Y ja:B2 Others Lingyue /ˀja:kDL1/ Saek /ja:lA2/ The above data show clearly that the reconstruction of PT *j- is dubious as most etyma are found only in SWT. Etyma that are found outside of SWT either show irregularities in the correspondence, or can be identified as loans. For example, the etymon ‘difficult’ is also found outside SWT, but the NT forms point to *ˀj- rather than *j-. As for ‘paternal grandmother’, it is found in both SWT and NT and clearly points to *j-. However, it most likely have a Mon-Khmer origin, cf. Proto-Mon-Khmer *yaʔ ‘grandmother’ (Shorto 2006). Therefore, Li’s *j- cannot be reconstructed for PT. 3.6.2.4 *ʔThe PT glottal stop is retained intact in all attested modern Tai varieties. Etyma that go back to this PT initial show tones in the first series, which indicates that this PT phoneme was voiceless. It is also reconstructed as *ʔ- in Li (1977: 243-250). Examples of PT etyma with *ʔ- are given in Table 3-26. 109 Table 3-26 Etyma with PT *ʔ- Gloss ‘to bathe’ ‘to open the mouth’ ‘satiated’ ‘chest’ ‘to exit’ PT *ʔa:pD *ʔa:C *ʔi:mB *ʔɤkD *ʔo:kD SI ʔa:pDL1 ʔa:C1 ʔimB1 ʔokDS1 ʔɔ:kDL1 LC ʔa:pDL1 ʔa:C1 ʔimB1 ʔɤkDS1 ʔo:kDL1 Y ʔa:pDL1 ʔa:C1 ʔimB1 ʔakDS1 ʔokDL1 Others 3.6.3 Voiced stops The PT voiced stops *b-, *d-, *ɟ-,*g-, and *ɢ- have been devoiced in most modern Tai varieties, except for a few dialects on the Sino-Vietnamese border. Dialects that have not gone through devoicing include Cao Bang, Wenma, and various Budai dialects (Haudricourt 1960; Kosaka 1997; L'Écôle Française d'Extême-Orient 1938; L-Thongkum 1997; Ross 1996; Zhang et al. 1999). For example, Cao Bang still retains voicing but has developed breathiness in all PT voiced stops. Therefore, PT *bis realized as [b̤ -] while PT *ɓ- is now realized as [b-] (Pittayaporn 2007d). Other dialects reflect PT voiced stops either as plain /p-/, /t-/, /c-/, and /k-/, or as aspirated /pʰ-/, /tʰ-/, /cʰ-/, and /kʰ-/. Etyma that go back to PT voiced stops have tones in the second series. 3.6.3.1 *bThis PT labial consonant has been devoiced in most Tai varieties. Most dialects including Shan, Black Tai, White Tai, Sapa, Lungming, Lungchow, Yay, Wuming, etc. have an unaspirated reflex /p-/, but many others such as Siamese, Lao, Southern Thai, Phuan, Nyo, Phu Thai, Leiping, and Saek show an aspirated stop /pʰ-/. 110 Cao Bang and Wenma still keep this sound as voiced /b̤ -/ and /b-/, respectively. Examples of etyma with PT *b- are provided in Table 3-27. Li (1977: 65-67) also reconstructs *b- for this set of etyma. Table 3-27 Etyma with PT *b- Gloss ‘expensive’ ‘blistered’ ‘to fold’ ‘father, man’ ‘to lay across’ PT *be:ŋA *bo:ŋA *bapD *bo:C *ba:tD SI LC pʰɛ:ŋA2 pe:ŋA2 pʰɔ:ŋA2 pʰapDS2 pʰɔ:B2 -t po:C2 pʰa:tDL2 Y peŋA2 poŋA2 papDS2 poC2 pa:tDL2 Others Cao Bang /b̤ ɛŋA2/ Cao Bang /b̤ apDS2/ Wenma /bɤwC2/ Cao Bang /b̤ a:tDL2/ The etymon ‘father’ has *B in SWT but *C elsewhere. The original tone was probably *C but SWT changed to *B due to contamination by *me:B ‘mother’ 3.6.3.2 *dPT *d- has become devoiced in most Tai varieties. Most dialects including Southern Shan, Black Tai, White Tai, Sapa, Lungming, Lungchow, Yay, Wuming, etc. have an unaspirated reflex /t-/, but many others such as Siamese, Lao, Southern Thai, Phuan, Nyo, Phu Thai, Leiping, and Saek show an aspirated stop /tʰ-/. Cao Bang, and Wenma still keep this sound as voiced /d̤ -/ and /d-/, respectively. Examples of etyma with PT *d- are provided in Table 3-28. Li (1977: 104-107) also reconstructs *d- for this set of data. 111 Table 3-28 Etyma with PT *d- Gloss ‘river’ ‘ashes38 ‘open field ‘to pound ‘stomach, belly’ PT SI *da:B tʰa:B2 *dawB tʰawB2 *doŋB tʰuŋB2 *dupD tʰupDS2 *dwu:ŋC tʰɔ:ŋC2 LC ta:B2 to:ŋC2 Y ta:B2 tawB2 toŋB2 tupDS2 tuŋC2 Others Wenma /doB2/ Wenma /dɤwB2/ Cao Bang /d̤ oŋB2/ Cao Bang /d̤ upDS2/ Wenma /duŋC2/ Note a semantic shift from ‘river’ to ‘pier’ in Siamese. 3.6.3.3 *ɟPT *ɟ- has become devoiced in most Tai varieties. A very common reflex of this PT sound is /c-/, which merged with the original *c-, as in Tai Yuan, Lue, White Tai, Lungming, Western Nung, Lungchow, Guangnan, and Yishan. In a few dialects, after merging with the original PT *c-, the secondary *c- from PT *ɟ- became affricated to /ts-/ as in Long’an, Debao, and Jingxi. In a number of dialects the affricate became a fricative /ɕ-/ as in Wuming, Tianlin, Lingyue, Donglan, Du’an, etc. Many CT and NT dialects including Shangsi, Chongzuo, and Daxin reflect this palatal stop *ɟ- as /s-/ possibly through an intermediate stage *z-. Cao Bang, which still retains original voicing, has /z-/, attests this intermediate step. In contrast, Wenma has a voiced affricate /dz-/ as the reflex of this PT voiced stop. In dialects that reflect voiced stops as aspirated, the voiced *ɟ- was also devoiced to *cʰ-. In Siamese, Southern Thai, and Leiping, it is reflected simply as /cʰ-/. In contrast, Lao, Phu Thai, 38 Li (1977: 95) cites Lunchow /pjawB2/ as the only example of *vl- but this form is in fact a different etymon— *brawB ‘leaf ashes’. Debao has both /tawB2/ from *dawB and /pjawB2/ from *brawB. 112 Phuan, Nyo etc. have changed the intermediate *cʰ- to /s-/. This PT onset corresponds to Li’s affricate *dʑ-39 (Li 1977: 168-173). Table 3-29 provides examples of PT *ɟ-. Table 3-29 Etyma with PT *ɟ- Gloss ‘to weigh’ ‘morning’ ‘to taste’ ‘name’ ‘rope’ PT *ɟaŋB *ɟa:wC *ɟi:mA *ɟɤ:B *ɟɤ:kD SI cʰaŋB2 cʰa:wC2 cʰimA2 cʰɯ:B2 cʰɯəkDL2 LC caŋB2 cawC2 cimA2 cɤ:kDL2 Y saŋB2 sawC2 simA2 soB2 sa:kDL2 Others Cao Bang /zaŋB2/ Cao Bang /zawC2/ Cao Bang /zimA2/ Lungming /cɤɰB2/ Cao Bang /zɯəkDL2/ 3.6.3.4 *gThis PT consonant follows the same pathway as other voiced stops. It became devoiced in most dialects, except for Cao Bang and Budai dialects. In Cao Bang, it is reflected as a breathy /g̈ -/. In most other varieties including Southern Shan, Dehong, Black Tai, White Tai, Lungchow, Debao, Yay, Qinzhou, Shangsi, Wuming, Longsheng, etc, its reflex is a voiceless unaspirated stop /k-/. In dialects like Siamese, Lao, Phuan, Phu Thai, and Leiping, the reflex of this voiced velar stop is aspirated /kʰ-/. This PT onset corresponds to Li’s *g- (1977: 198-203). Table 3-30 presents examples of etyma with PT *g-. 39 *ǰ- in Li’s notation. 113 Table 3-30 Etyma with PT *g- Gloss ‘stuck’ ‘handle, rod’ ‘narrow, tight’ ‘pair’ ‘pliers, thongs’ PT *ga:A *galA *gapD *gu:B *gi:mA SI kʰa:A2 kʰanA2 kʰapDS2 kʰu:B2 kʰi:mA2 LC ka:A2 kapDS2 ku:B2 kimA2 Y ka:A2 kanA2 kuB2 cimA2 Others Wenma /goA2/ Cao Bang /g̈ apDS2/ Wenma /guB2/ Note that the initial /c-/ in the Yay form for ‘tongs’ is a result of regular palatalization before front vowels. 3.6.3.5 *ɢThe voiced uvular stop is reflected in a number of ways in modern languages. In SWT and most CT dialects, it merged with the velar *g-, which gives /k-/ or /kʰ-/ depending on whether voiced stops are reflected as plain or aspirated. Black Tai, White Tai, Lungming, Lungchow, Debao, etc. all have /k-/. On the other hand, Siamese, Lao, Nyo, Phuan, Phu Thai, etc. show /kʰ-/. In NT as well as a few CT dialects, PT *ɢ- was spirantized and merged with PT *ɣ-. The modern reflex of the merger is /kʰ-/ in Shangsi and Fusui; /x-/ in Wuming; /h-/ in Yay, Hechi, Tiandong, Longsheng, etc.; /ɣ-/ in Saek and Qiubei; and /g-/ in Huanjiang. Etyma with *ɢ- are included under *ɣ- in Li (1977: 214-219). See §3.3 for discussion of the reconstruction of uvular consonant. Table 3-31 presents examples of etyma that go back to PT *ɢ-. 114 Table 3-31 Etyma with PT *ɢ- Gloss ‘carrying pole’ ‘chin, jaw’ PT *ɢa:nA *ɢa:ŋA SI kʰa:nA2 kʰa:ŋA2 LC ka:nA2 ka:ŋA2 Y ha:nA2 ha:ŋA2 Others Saek /ɣa:nA2/ Saek /ɣa:ŋA2/ It is interesting that only a few etyma with PT *ɢ- as onset can be reconstructed. 3.6.4 Voiceless fricatives There were altogether five voiceless fricatives that can function as onsets in PT: *s-, *ɕ-, *x-, *χ- and *h-. Except for *h-, the reflexes of these fricatives in modern dialects are usually different from their original values. The etyma that go back to these PT voiceless fricatives all have tones in the first series. 3.6.4.1 *sThe PT voiceless alveolar fricative *s- shows a wide range of reflexes in modern varieties. It is preserved as /s-/ in most SWT dialects including Siamese, Lao, Lue, Black Tai, White Tai, etc.; in CT varieties such as Lungming, Western Nung, Jingxi, Yishan, etc.; and in NT dialects including Longsheng, Laibin, and most Bouyei dialects. It is fronted to /θ-/ in many NT varieties including Wuming, Hengxian, Liujiang, Lianshan, etc., and in the CT dialects of Debao and Jingxi. Another common reflex of *s- is the lateral /ɬ-/, which is found generally in CT dialects, including Qinzhou, Shangsi, Ningming, Lungchow, etc., as well as a few NT dialect such as Yongbei and Tiandong. In Bao Yen, the lateral /ɬ-/ has become an aspirated stop /tʰ-/. A special development is found in Southern Shan, where PT *s- is now reflected as 115 aspirated /sʰ-/ in contrast with /s-/ which comes from PT *c-. The distinction between PT *s- and *c- is thus still preserved. Another special development is the case of Hechi and Huanjiang, where PT *s- is regularly reflected as /r-/. Li (1977:152-161) also reconstructs this phoneme as *s-. Examples of PT *s- is given in Table 3-32. Table 3-32 Etyma with PT *s- Gloss ‘unmarried woman’ ‘intestine’ ‘heel’ ‘four’ ‘to order’ PT *sa:wA *sajC *sɤnC *si:B *saŋB SI sa:wA1 sajC1 sonC1 si:B1 saŋB1 LC ɬa:wA1 ɬajC1 ɬɤnC1 ɬi:B1 ɬaŋB1 Y θa:wA1 θajC1 θanB1 θi:B1 θaŋB1 Others Bao Yen /tʰa:wA1/ Bao Yen /tʰɤjC1/ Bao Yen /tʰɤnC1/ Bao Yen /tʰi:B1/ Bao Yen /tʰaŋB1/ 3.6.4.2 *ɕThis voiceless palatal fricative is tentatively reconstructed to account for a discrepancy between SWT/CT on one hand and NT on the other. The only example is ‘owner’ which is reflected as /ca:wC1/ in Siamese and /cawC1/ in Lungchow, but /θuC1/ in Yay and /su:C1/ in Saek. However, there are parallel cases in the voiced series (see §3.6.5.2). In this case, SWT/CT dialects point to earlier *c- (see §3.6.1.3) while their NT counterparts point to *s- (see §3.6.4.1). In this account, *ɕ- strengthened to *c- in SWT/CT but merged with *s- in NT. Li (Li 1977: 165, 167) includes this etymon under *tɕ-40. 40 *č- in Li’s notation 116 3.6.4.3 *xThe PT voiceless velar fricative is rarely preserved unchanged in modern dialects. It merged with PT *h- in all NT dialects including Yay, Wuming, Po-ai, Longsheng, Rong’an, etc. as well as some CT varieties such as Fusui, Shangsi, and Long’an. It is now reflected a /h-/ in these dialects. In most SWT and CT varieties, in contrast, it merged with secondary *kʰ- which came either from voicing alternation items (see §3.4), aspiration of *kr- (see §3.5.2), or borrowings (see §3.5.1). For instance, /kʰ-/ in Siamese came from PT *x- (> PSWT *x-), PT *q- (> PSWT *q-), PT *C.k- (> PSWT *kʰ-), *C.q- (> PSWT *kʰ-), PT kr- (> PSWT *kʰr-), or PSWT borrowing (e.g. PSWT *kʰajA ‘open’ from 開 kāi). Dehong and many other dialects in the area have /x-/ for *x-, but this /x-/ is an equivalent of Siamese /kʰ-/, an end result of multiple mergers. Only in White Tai and some Lue dialects (Li 1964; 1977: 192, 207) is the velar fricative still retained as a distinct phoneme. Li (1977: 207-214) listed etyma with this PT fricative under *x-. Table 3-33 gives examples of etyma with PT *x-. Table 3-33 Etyma with PT *x- Gloss ‘green’ ‘galangal ‘white’ ‘hook’ PT *xiəwA *xa:B *xa:wA *xo:A SI kʰiəwA1 kʰa:B1 kʰa:wA1 kʰɔ:A1 LC Y kʰe:wA1 hewA1 ha:B1 kʰa:wA1 ha:wA1 Others White Tai /xɛwB1/ White Tai /xa:B1/ White Tai /xa:wA1/ White Tai /xɔA1/ Note the additional examples for ‘hook’: Debao /kʰo:A1/, Lungming /kʰo:A1/, Saek /hɔ:A1/, and Lingyue /hɔA1/. They all point to PT *x-. 117 3.6.4.4 *χThe voiceless uvular fricative *χ- merged with *x- in most dialects. Therefore, reflexes of *χ- is generally identical to those of *x- (see §3.3 and §3.6.4.3). Exceptions are Phuan and Phu Thai dialects (2008b; Pittayaporn to appear-b), as well as some SWT dialects of central Vietnam (1999; Ferlus 2008)41. In these dialects, PT *χmerged with *h- (see §3.3) rather than *x-. Li (1977: 207-214) includes this set of etyma under *x-. Examples of PT *χ- are provided in Table 3-34. Table 3-34 Etyma with PT *χ- Gloss ‘to cross’ ‘to crow’ ‘fang’ ‘to go up’ ‘ginger’ PT *χa:mC *χalA *χe:wC *χɯ:nC *χi:ŋA SI kʰa:mC1 kʰanA1 kʰiəwC1 kʰɯnC1 kʰiŋA1 LC kʰa:mC1 kʰanA1 kʰe:wC1 kʰɤnC1 kʰiŋA1 Y ha:mC1 hanA1 hewC1 hɯnC1 hiŋA1 Others Kapong /ha:mC1/ Kapong /hanA1/ Kapong /hɛ:wC1/ Kapong /hɯ:nC1/ Kapong /hi:ŋA1/ Data from the Kapong dialect of Phu Thai is given above. In this variety, *x- is reflected as aspirated stop /kʰ-/ but *χ- is reflected as /h-/. Contrast /hɛ:wC1/ ‘tooth’ from PT *χe:wC ‘fang’ and /kʰiəwA1/ ‘green’ from PT *xiəwA. 3.6.4.5 *hThe voiceless glottal fricative is retained in almost all modern dialects. There are a few varieties where it is reflected as /x-/, /ɣ-/, or /ɦ-/. It is reflected as /x-/ in 41 Data from the SWT varieties in central Vietnam were not systematically examined in this study. Therefore, it does not appear on the map in Figure 1-3. 118 Wenma and Yishan, as /ɣ-/ in Qiubei and some Bouyei dialects (points 14, 18, 22, 23, 24, etc. in Bùyīyǔ Diàochá Bàogào), and as /ɦ-/ in some other Bouyei dialects (point 7, 15, 19, etc.). It is also reconstructed as *h- by Li (1977: 249-253). Etyma in Table 3-35 are examples of PT *h-. Table 3-35 Etyma with PT *h- Gloss ‘five’ ‘goose’ ‘to give’ ‘fragrant’ ‘to cover with cloth’ PT SI *ha:C ha:C1 *ha:nB ha:nB1 *hajC hajC1 *ho:mA hɔ:mA1 *hɤmB homB1 LC ha:C1 ha:nB1 hɯ:C1 ho:mA1 humB1 Y ha:C1 ha:nB1 haɰC1 homA1 Others Bouyei(7) /ɦaC1/ Bouyei(7) /ɦanC1/ Bouyei(7) /ɦaɰC1/ Saek /hamB1/ 3.6.5 Voiced fricatives There were perhaps only three voiced fricatives that can function as onsets in PT: *z-, *ʑ-, and *ɣ-. There is no evidence for voiced fricatives at the uvular and glottal places of articulation, the two places farthest back in the vocal tract. When these voiced fricatives were devoiced, they generally merged with their voiceless counterparts *s-, *ɕ-, and *x-, respectively. The etyma that go back to these PT voiced fricatives all have tones in the second series. 3.6.5.1 *zThe PT voiced alveolar fricative *z- shows reflexes identical with those of PT *s- in most dialects. This is because *z- became *s- after devoicing of original voiced 119 consonants. Cao Bang and Saek are two special cases. In Cao Bang, *z- became rhoticized to /r-/, while the original *ɟ- became /z-/ to fill in the gap. Recall that Cao Bang retains the original voicing of PT obstruents. In Saek, *z- became /j-/ merging with PT *ˀj-, probably due to influence from North-Central Vietnamese, where original *z- has become /j-/ (Alves 2007)42. This PT fricative corresponds to Li’s *z(1977: 161-164). Table 3-36 gives examples of PT *z-. Table 3-36 Etyma with PT *z- Gloss ‘to wash (clothes’ ‘straight’ ‘fish trap’ ‘to repeat’ PT *zakD *zɤ:B *zajA *zamC SI sakDS2 sɯ:B2 sajA2 samC2 LC ɬakDS2 ɬɯ:B2 ɬajA2 Y θakDS2 θoB2 θajA2 Others Saek /jakDS2/ Saek /jo:B2/ Saek /jajA2/ Saek /jamC2/ 3.6.5.2 *ʑThe PT voiced palatal fricative is tentatively reconstructed to account for a discrepancy between SWT/CT on one hand and NT on the other. For etyma in Table 3-37, SWT/CT dialects point to earlier *ɟ- (see §3.6.1.3) while their NT counterparts point to *z- (see §3.6.5.1). In this account, *ʑ- strengthened and merged with *ɟ- in SWT/CT but merged with *z- in NT. Li (1977: 165, 167) includes these etyma under *dʑ-43. Also, refer to the reconstruction of *ɕ- in §3.6.4.2. 42 According to Chamberlain (2000), Saek speakers in northeastern Thailand and Central Laos migrated from modern-day Thanh Hoa, Nghe An and Quang Binh Provinces of Vietnam. Also see Saek tales recorded by Gedney (Hudak 1993). 43 *ǰ- in Li’s notation 120 Table 3-37 Etyma with PT *ʑ- Gloss ‘man, male’ ‘eaves’ ‘to commission’ PT *ʑa:jA *ʑa:jA *ʑaɰC SI cʰa:jA2 cʰa:jA2 cʰajC2 LC ca:jA2 ca:jA2 caɰC2 Y θa:jA2 θa:jA2 θaɰC2 Others SaekC2 /ja:jA2/ For ‘man’ and ‘to commission’, Saek has /sa:jA2/ and /sajC2/ respectively. The onset /s-/ is unexpected because we would expect Saek to show /j-/ (from *z- < PT *ʑ). However, both cases can be viewed as influence from Lao or Siamese. The Lao forms for these two etyma are /sa:jA2/ and /sajC2/ respectively. The loan nature of the Saek form for the latter is especially obvious as the expected Saek vowel for this etymon is /ɤ:/, not /aj/. 3.6.5.3 *ɣThe PT voiced velar fricative generally merged with its voiceless counterpart after the devoicing of original voiced onsets (see §3.6.4.3). Only a few languages show special developments. Saek is conservative in that it retains PT *ɣ- intact. Black Tai and Lungchow reflect *x- as /kʰ-/ but regularly hardened *ɣ- to *g- before devoicing the merged phoneme to /k-/. Wuming merged PT *x- with *h- but preserved PT *ɣ- until it became /x-/ after the devoicing. Many CT dialects such as Leiping, Lungming, and Western Nung now have /h-/ for PT *ɣ- but /kʰ-/ for *x-. Debao is unique in having sonorant reflexes for PT *ɣ-. Now this PT fricative is reflected as /v-/ (from an intermediate *w-) before front vowels but /j-/ elsewhere. This PT fricative corresponds to Li’s *ɣ- (Li 1977: 214-219), but refer to the discussion of PT dorsal consonants in 3.3. Table 3-38 gives examples of PT * ɣ-. 121 Table 3-38 Etyma with PT *ɣ- Gloss ‘night time’ ‘night’ ‘neck’ ‘dike between rice fields’ ‘thatch grass’ PT SI LC *ɣamB kʰamB2 kamB2 *ɣɯ:nA kʰɯ:nA2 kɤnA2 *ɣo:A kʰɔ:A2 ko:A2 *ɣalA kʰanA2 kanA2 *ɣa:A kʰa:A2 ka:A2 Y hamB2 hɯnA2 hoA2 hanA2 haA2 Others Saek /ɣamB2/ Saek /ɣɯnA2/ Saek /ɣɔ:A2/ Saek /ɣalA2/ Saek /ɣa:A2/ 3.6.6 Voiced nasals There were altogether four plain voiced nasals that could function as onsets in PT: *m-, *n-, *ɲ-, and *ŋ-. After their voiceless counterparts became voiced, the newly created voiced nasals merged with these original plain nasals. However, etyma that go back to PT *m-, *n-, *ɲ-, and *ŋ- are generally reflected with tones in the second series while the voiceless ones are reflected with tones in the first series. 3.6.6.1 *mWithout exception, the plain nasal *m- is retained in all modern dialects. Now PT *m- is reflected as /m-/ with tones in the second series in all modern dialects. Li (1977: 71-73) also reconstructs *m- for this set of data. Table 3-39 presents example of PT *m-. 122 Table 3-39 Etyma with PT *m- Gloss ‘mother, woman’ ‘ant’ ‘yam’ ‘mucus’ PT *me:B *mɤcD *manA *mu:kD SI mɛ:B2 motDS2 manA2 mu:kDL2 LC me:B2 mɤtDS2 manA2 mukDS2 Y meB2 matDS2 manA2 mukDS2 Others 3.6.6.2 *nThe plain alveolar nasal *n- is preserved in most modern dialects. Only in a few Shan varieties did it merge completely with PT *l-. For example, in the Mengshi and Chefang dialects of Tai Nüa, PT *n- is reflected as /l-/. Etyma with PT *n- now show tones in the second series. This alveolar nasal corresponds to Li’s *n- (Li 1977: 111-113). Table 3-40 presents example of PT *n-. Table 3-40 Etyma with PT *n- Gloss ‘paddy field’ ‘otter’ ‘rotten’ ‘slow’ ‘small’ PT *na:A *na:kD *nawB *na:nA *no:jC SI na:A2 na:kDL2 nawB2 na:nA2 nɔ:jC2 LC na:A2 na:kDL2 no:jC2 Y na:A2 na:kDL2 nawB2 na:nA2 Others Mengshi /la:A2/ Mengshi /lawB2/ The etymon ‘small’ shows a tonal irregularity, probably due to confusion with *ʰno:jB ‘few, little’. Lungming has /no:jC2/ for ‘minor’, and /no:jB1/ for ‘few, little’. 123 Similarly, in Siamese /nɔ:jC2/ occurs with /lekDS2/ ‘small’ in the co-ordinate compound /lekDS2 nɔ:jC2/ ‘small (size, degree)’ but /nɔjB1/ occurs with /nitDS2/ ‘litte (amount) in the expression /nitDS2 nɔjB1/ ‘small (amount)’. 3.6.6.3 *ɲIn modern Tai languages, the plain palatal nasal *ɲ- is either preserved as /ɲ-/ or denasalized to a palatal glide /j-/. While Lao, Black Tai, White Tai, Western Nung, Debao, Yishan, Tianlin, Huanjiang, Guigang, Saek, etc. show the conservative reflex /ɲ-/, Siamese, Lue, Leiping, Lungchow, Qinzhou, Shangsi, Wuming, Hengxian, Tiandong etc. show the glide reflex /j-/. This palatal nasal corresponds to Li’s *ɲ- (Li 1977: 173-176). Table 3-40 presents example of PT *ɲ-. Table 3-41 Etyma with PT *ɲ- Gloss PT ‘to sew’ *ɲɤpD ‘to shoot’ *ɲɯŋA2 ‘tip, highest point’ *ɲo:tD ‘mosquito’ *ɲuŋA SI jepDS2 jiŋA2 jɔ:tDL2 juŋA2 LC japDS2 jaŋA2 jo:tDL2 juŋA2 Y ɲipDS2 ɲiŋA2 ɲuŋA2 Others Lao /ɲipDS2/ Lao /ɲiŋA2/ Lao /ɲɔ:tDL2/ Lao /ɲuŋA2/ 3.6.6.4 *ŋThe plain velar nasal *ŋ- is preserved in almost all modern dialects. The only exceptions are Saek and certain Southern Thai dialects. While PT *ŋ- became /ɲ-/ in Saek, it became /h-/ with tones in the second series in some dialects of Southern Thai. This velar nasal corresponds to Li’s *ŋ- (Li 1977: 205-206). Table 3-42 presents example of PT *ŋ-. 124 Table 3-42 Etyma with PT *ŋ- Gloss ‘early meal’ ‘easy’ ‘crocodile’ ‘elephant tusk’ ‘shadow’ PT *ŋa:jA *ŋa:jB *ŋɯəkD *ŋa:A *ŋawA SI LC ŋa:jA2 ŋa:jA2 ŋa:jB2 ŋɯəkDL2 ŋɤ:kDL2 ŋa:A2 ŋawA2 ŋawA2 Y ŋa:jA2 ŋa:jB2 ŋɯəkDL2 ŋa:A2 ŋawA2 Others Note a semantic shift from ‘crocodile’ to ‘snake-like mythical water creature’ in a great number of SWT and CT dialects, including Siamese, Lao, Black Tai, Tai Nüa, Lungming, Lungchow, Bao Yen, Sapa etc. Many CT dialects in Guangxi and all NT varieties except Saek still preserve the original meaning. The Chinese source for this etymon 鱷 è (MC ŋâk) also means ‘crocodile’ (Pulleyblank 1991: 87; Schuessler 2007: 223). 3.6.7 Voiceless nasals In addition to the plain nasals, PT had four voiceless nasals *ʰm-, *ʰn-, *ʰɲ-, and *ʰŋ-. These PT phonemes generally merged with their respective voiced counterparts so that etyma with these PT voiceless nasals have tones in the first series in modern dialects (see §6.2). Chronologically, the voicing of the voiceless nasals must have preceded the devoicing of voiced obstruents because of all the dialects that still preserve original voiced obstruents have voiced their sonorants. L-Thongkum (1997) argues convincingly that it is this voicing of voiceless sonorants that caused the subphonemic pitch difference to become a binary register split. 125 3.6.7.1 *ʰmThe PT voiceless labial nasal *ʰm- fell together with its voiced counterpart *m- . Therefore, its reflex in each dialect is identical to that of PT *m- (see §3.6.6.1). This labial sonorant corresponds to Li’s *ʰm- (Li 1977: 74-76). Table 3-43 presents example of PT *ʰm-. Table 3-43 Etyma with PT *ʰm- Gloss ‘dog’ ‘pig’ ‘pot’ ‘new’ ‘soot’ PT *ʰma:A *ʰmu:A *ʰmo:C *ʰmɤ:lB *ʰmi:C SI ma:A1 mu:A1 mɔ:C1 majB1 mi:C1 LC ma:A1 mu:A1 mo:C1 maɰB1 mi:C1 Y ma:A1 mu:A1 moC1 moB1 miC1 Others 3.6.7.2 *ʰnThe PT voiceless alveolar nasal *ʰn- merged with its plain counterpart *n-. Therefore, its reflex in each dialect is identical to that of PT *n- (see §3.6.6.2). This alveolar nasal is also reconstructed by Li (Li 1977: 113-116) as *ʰn- Table 3-44 provides example of PT *ʰn-. 126 Table 3-44 Etyma with PT *ʰn- Gloss ‘heavy’ ‘to steam’ ‘face’ ‘skin’ ‘mouse’ PT *ʰnakD *ʰnɯŋC *ʰna:C *ʰnaŋA *ʰnu:A SI nakDS1 nɯŋC1 na:C1 naŋA1 nu:A1 LC nakDS1 na:C1 naŋA1 nu:A1 Y nakDS1 naŋC1 na:C1 naŋA1 nu:A1 Others Mengshi /lakDS1/ Mengshi /lɯŋC1/ Mengshi /laC2/ Mengshi /laŋA1/ Mengshi /luA1/ 3.6.7.3 *ʰɲThe PT voiceless palatal nasal *ʰɲ- fell together with its voiced counterpart *ɲ-. Therefore, its reflex in each dialect is identical to that of PT *ɲ- (see §3.6.6.1). This palatal sonorant corresponds to Li’s *ʰɲ- (Li 1977: 176-178). Table 3-45 presents example of PT *ʰɲ-. Table 3-45 Etyma with PT *ʰɲ- Gloss ‘grass’ ‘coarse, tough’ ‘to step on’ ‘to stretch out’ PT *ʰɲɯəC *ʰɲa:pD *ɲamB *ʰɲiətD SI ja:C1 ja:pDL1 jamB2 jiətDL1 LC ja:C1 ja:pDL1 jamB2 ji:tDL1 Y ɲiəC1 ɲa:pDL1 ɲamB2 ɲiətDL1 Others Lao /ɲa:C1/ Lao /ɲa:pDL1/ Lao /ɲamB2/ 3.6.7.4 *ʰŋThe PT voiceless velar *ʰŋ- has two main reflexes in modern Tai dialects: /ŋ-/ and /h-/. Many varieties simply voiced the nasal *ʰŋ- to /ŋ-/ merging with the original 127 *ŋ-. These dialects include Siamese, Lao, Lue, Western Nung, Debao, Long’an, Guangnan Nung, Tiandong, Tianlin, Lingyue, etc. In contrast, many dialects merged *ʰŋ- with *h- before the voicing of voiceless sonorants took place. Dialects in this group include Black Tai, White Tai, Leiping, Lungchow, Shangsi, Ningming, Chongzuo, Saek, etc. This voiceless velar nasal corresponds to Li’s *ʰŋ- (Li 1977: 206-207). Table 3-46 presents example of PT *ʰŋ-. Table 3-46 Etyma with PT *ʰŋ- Gloss ‘gum, gill’ ‘gray (hair)’ ‘to lie face up’ ‘to tilt head up’ PT *ʰŋɯək *ʰŋo:kD *ʰŋa:jA *ʰŋe:nA SI LC ŋɯəkDL1 hɤ:kDL1 ŋɔ:kDL1 ŋa:jA1 ŋɛ:nA1 Y Others ŋɯkDS1-t Saek /hu:kDL1/ ha:jA1 Because examples of PT *ʰŋ- are very rare, its reconstruction is still tentative. It is possible that the set of etyma assigned to PT *ʰŋ- are secondary, as only a few examples exist, some of which are only found in SWT. The regular reflex in Yay is perhaps /h-/; the irregularity in tone in the etymon ‘gill, gum’ indicates that it might be secondary, i.e. dialect borrowing or late loan from a Chinese dialect, cf. 顎 è (< MC ŋâk, OC *ŋâk) ‘palate’. Perhaps the true PT *ʰŋ- merged with PT *h- early on as suggested by the onset *h- in many PT etyma that go back to *ŋ- (perhaps from earlier *ʰŋ-) in Chinese, e.g. PT *ha:C from 五 wǔ (< MC ŋuoB, OC *ŋâʔ) ‘five’, and PT *ha:nB from 雁 (< MC ŋanC, OC *ŋrâns) ‘wild goose’ 128 3.6.8 Voiced liquids and glides There were two PT voiced liquids and one voiced glide in PT: *r-, *l-, and *w-. Each of them has a voiceless counterpart with which they usually merged. However, these plain voiced liquids and glides are generally reflected with tones in the second series while the voiceless ones are reflected with tones in the first series. 3.6.8.1 *rThe PT voiced alveolar trill *r- shows a wide range of reflexes in modern dialects, including /r-/, /l-/, /h-/, /ʰl-/, /ɬ-/, /ʰj-/, /j-/, /z-/ and /ð-/. Many languages from the three clusters have a conservative reflex /r-/, including Siamese, Southern Tai, Bao Yen, Cao Bang, Jingxi, Yishan, Yay, Saek, Huanjiang, Longsheng, and some dialects of Wuming. Most SWT dialects aspirated the trill into /h-/, including Lao, Phuan, Phu Thai, Tai Yuan, Lue, Aiton, White Tai, Sapa, etc. In many CT and NT dialects, *rbecame a lateral /l-/, as in Lungming, Debao, Qinzhou, Chongzuo, Shangsi, Yongbei, Lingyue, and Donglan, to cite a few. A few varieties attest /ʰl-/ from PT *r-, which is kept apart from original *l- which is reflected as plain /l-/ (see §3.6.8.2). Dialects in this group include Yongnan and Long’an. Some dialects in the NT group including Liujiang, Yishan, and Shanglin yodicized PT *r- to /j-/. A few also have a yodicized reflex but with pre-aspiration /ʰj-/, distinct from /j-/. Many dialects, mostly in the NT group, e.g. Western Nung, Hengxian, Qiubei, and Guangnan Sha, show an interdental fricative /ð-/ as the reflex of the PT trill. Rong’an went one step further and devoiced its intermediate *ð- to /θ-/. Another group of dialects merged PT *r- with *z-, which is now reflected as /ɬ-/ (see §3.6.5.1). Dialects in this group include the CT dialects of Leiping, Daxin, Lungchow, etc. The intermediate step is attested in Wenma, which still retains /z-/. This PT phoneme corresponds to Li’s *r- (Li 1977: 142-148). Table 3-47 presents example of PT *r-. 129 Table 3-47 Etyma with PT *r- Gloss ‘house’ ‘hundred, to string’ ‘to know’ ‘long’ ‘strength’ PT SI LC *rɤ:nA rɯənA2 ɬɤ:nA2 *ro:jC rɔ:jC2 ɬo:jC2 *ru:wC ru:C2 ɬu:C2 *rɯjA2 ri:A2 ɬi:A2 *re:ŋA2 rɛ:ŋA2 ɬe:ŋA2 Y ra:nA2 rojC2 roC2 rajA2 reŋA2 Others W. Nung /ðɯnA2/ W. Nung /ðojC2/ W. Nung /ðu:C2/ W. Nung /ði:A2/ W. Nung /ðeŋA2/ Note a semantic shift from ‘long’ to ‘oval’ in Siamese. 3.6.8.2 *lThe PT voiced lateral *l- is retained as /l-/ in all attested Tai dialects. Li (1977: 133-136) also reconstructs *l- for this set of data. Table 3-48 presents example of PT *l-. Table 3-48 Etyma with PT *l- Gloss ‘child (offspring)’ ‘blood’ ‘monkey’ ‘to fall’ ‘to forget’ PT SI LC *lɯ:kD lu:kDL2 lukDL2 *lɯətD lɯətDL2 lɤ:tDL2 *li:ŋA liŋA2 liŋA2 *lɤmC lomC2 *lɯ:mA lɯ:mA2 lumA2 Y Others lɯkDS2 lɯətDL2 liŋA2 lamC2 Debao /lamC2/ lumA2 130 3.6.8.3 *wThe PT labial glide has various reflexes in modern dialects but they can be considered end results of two main lines of development. The first line is an innovative one in which *w- became a fricative *v-. This intermediate step is attested in Cao Bang and Wenma as /v-/. Most dialects underwent further devoiced this *v- to /f-/ when devoicing occurred. These dialects include Siamese, Black Tai, White Tai, Lao, Lungchow, Lungming, Shangsi, Qinzhou, Debao, Yay, Pinguo, Liujiang, Rong’an, etc. However, in many Shan varieties including Southern Shan, Phake, and Aiton, the /f-/ has become /pʰ-/. Another line of development is attested in Qiubei and Saek. These two varieties did not go through the early change from *w- to *v-. Qiubei is more conservative because the PT labial glide is kept as /w-/. Saek deserves special attention. Although PT *w- is now reflected as /v-/, this voiced fricative is not a retention of the intermediate *v- of the other line of development above. In fact, PT *w- was kept as intact in Saek at least until the devoicing, when it merged with PT *ʰw- (see §3.6.9.3). The result of the merger between PT *w- and *ʰw- was *w-. It is this merged *w- that became /v-/ in Saek. This change from *w- to /v-/ is a much later change than the change from PT *w- to *v- in discussed above. Table 3-49 presents examples of PT *w-. 131 Table 3-49 Etyma with PT *w- Gloss ‘fire’ ‘tooth’ ‘swollen’ ‘firewood’ ‘straw’ PT SI *wi:A fajA2 *wanA fanA2 *wokD fokDS2 *wɯ:nA fɯ:nA2 *wɯəŋA fa:ŋA2 LC Y Others fajA2 fi:A2 Qiubei /weiA2/ fanA2 fanA2 fokDS2 Qiubei /wokDS2/ fɤnA2 funA2 Qiubei /wənA2/ fɯəŋA2 fɯəŋA2 Qiubei /ɣuŋA2/ These etyma also have *w- in Proto-Lakkja (L-Thongkum 1992), e.g. *wanA ‘tooth’, and *wokD ‘swollen’. This PT labial glide corresponds to Li’s *v- (Li 1977: 79-81). What he reconstructs as *w- (Li 1977: 81-82) is in fact not reconstructible for PT as there are very few etyma in this set. Crucially, those etyma do not have regular correspondence, or can be shown to be borrowings. Table 3-50 gives examples of etyma that Li assigns to *w-. Table 3-50 Some etyma that Li reconstructs with *w- Gloss ‘to say’ ‘to place’ ‘deep pool’ ‘work’ ‘free, idle’ Siamese wa:B2 wajC2 waŋA2 wa:ŋB2 -t Lungchow Yay va:B2 vajC2 vaŋA2 viəkDL2 viəkDL2 va:ŋB1 Others Longsheng /va:ŋB1/ 132 Among these, only two are found in NT. However, it is exactly those two that can be identified as loans from Chinese, i.e ‘work’ from 役 yì (< MC jiwäk) ‘corvée duty’ perhaps through Vietnamese việc, and ‘free, idle’ possibly from 亡 wáng (< MC mjwaŋ) ‘not present’ or 罔 wǎng (< MC mjwaŋB) ‘not have’. The etymon ‘deep pool’ is also clearly a Chinese loan from, i.e. 汪 wāng (< MC ʔwâŋ) ‘deep pool’. Further, the etymon ‘to say’ may also be a Chinese loan, cf. 話 huà (< MC ɣwaiC) ‘speech’. 3.6.9 Voiceless liquids and glides There were two voiceless liquids and one voiceless glide in PT: *ʰr-, *ʰl-, and *ʰw-. Each of them has a voiced counterpart with which they usually merged. However, these voiceless liquids and glides are generally reflected with tones in the first series while the voiced ones are reflected with tones in the second series. 3.6.9.1 *ʰrIn most dialects, the PT voiceless trill *ʰr- fell together with its voiced counterpart *r-. Therefore, its reflex is identical to that of *r- (see §3.6.8.1). However, a number of dialects including Siamese, Lao, Lungchow, and Western Nung, underwent a change from *ʰr- to *h-, which occurred before PT voicedless sonorants became voiced. This PT trill corresponds to Li’s *ʰr- (Li 1977: 148-151). Table 3-51 presents example of PT *ʰr-. 133 Table 3-51 Etyma with PT *ʰr- Gloss ‘to yawn’ ‘to bark’ ‘dried up’ ‘withered’ ‘to shrink’ PT *ʰra:wA *ʰrawB *ʰre:ŋC *ʰriəwB *ʰrotD SI ha:wA1 hawB1 hɛ:ŋC1 hiəwB1 hotDS1 LC ha:wA1 hawB1 he:wB1 Y ra:wA1 rawB1 reŋC1 rewB1 rutDS1 Others W. Nung /hɛŋC1/ W. Nung /hotDS1/ 3.6.9.2 *ʰlThe PT voiceless lateral *ʰl- fell together with its voiced counterpart *l-. Therefore, its reflex in each dialect is identical to that of *l- (see §3.6.8.2). This PT lateral corresponds to Li’s *ʰl- (Li 1977: 136-141). Table 3-52 presents example of PT *ʰl-. Table 3-52 Etyma with PT *ʰl- Gloss ‘many, much’ ‘liquor’ ‘sharp-pointed’ ‘to be left over’ ‘to fall off’ PT SI *ʰla:jA la:jA1 *ʰlawC lawC1 *ʰle:mA lɛ:mA1 *ʰlɯəA lɯəA1 *ʰlɤnB lonB1 LC la:jA1 lawC1 le:mA1 lɯ:A1 Y la:jA1 lawC1 lemA1 lɯəA1 lanB1 Others Lungming /lɤnB1/ 3.6.9.3 *ʰwParallel to PT *w- (see 3.6.8.3), PT *ʰw- went through three lines of development. The first line of development is characterized by an early change from *ʰw- to *f-. This intermediate *f- has two main reflexes. The first reflex is /f-/ as found in Siamese, Lao, 134 Lue, Black Tai, Wuming, Pingguo, Tiandong, etc. The second reflex is /pʰ-/. This further change from *f- to /pʰ-/ is found in Aiton, Phake, etc. CT dialects, in general, show the second line of development. In Lungchow, Debao, Lungming, Shangsi, Daxin, Jingxi, etc., the reflex of *w- is /f-/ but the reflex of PT *ʰw- is /pʰ-/. This indicates that *ʰw- did not become *w- when the original voiceless sonorants became voiced. Instead, PT *ʰwmust have developed into /pʰ-/ before the voicing took place. Another line of development is represented by many NT dialects including Qiubei, Tianlin, Guangnan, Lingyue, Longsheng, Du’an, Shanglin, etc. In these dialects PT *w- is retained as /w-/. Liujiang, Yishan, Laibin and a few other dialects support an original value *ʰw- as they preserve the pre-aspiration in the form of /h-/ in etyma with PT nucleus /ɯə/, e.g. Liujiang /huC1/ ‘cloud’ from *ʰwɯəC < PT *fɯəC. In addition, Saek, Yay, and many Bouyei dialects have gone one step further and turned /w-/ into a fricative /v-/. This onset corresponds to Li’s *f- (Li 1977: 77-80). Examples of etyma with PT *ʰw- are given in Table 3-53. Table 3-53 Etyma with PT *ʰw- Gloss ‘lid’ ‘cloud’ ‘dam’ ‘astringent’ ‘soft-shelled turtle’44 PT SI *ʰwa:A fa:A1 *ʰwɯəC fa:C1 *ʰwa:jC fa:jA1 *ʰwɯətB fa:tDL1 *ʰwɯəA fa:A1 LC pʰa:A1 pʰa:C1 pʰa:jA1 pʰa:A1 Y va:A1 vɯət Others Liujiang /wa:A1/ Liujiang /huC1/ Liujiang /wa:jA1/ Liujiang /hwɤA1/ 44 It is not clear why the rimes in Liujiang forms for ‘cloud’ and ‘soft-shelled turtle’ are different. 135 Zhang (1995) proposes that this phoneme should be reconstructed as *pw- at the PT level. This is probably not the case because Guigang and nearby dialects, which generally show /w-/ for this PT sound, have /f-/ in front of /ɯə/ as in ‘cloud’ and ‘softshelled turtled’. If the original PT sound was *pw-, we would expect to see /p-/ in that environment. Rather, the /f-/ in Guigang forms for ‘cloud’ and ‘soft-shelled’ should be considered a special development of *ʰw- before /ɯ/. The Proto-Lakkja (L-Thongkum 1992) correspondent of ‘cloud’ lends further support to this interpretation, cf. ProtoLakkja *ʰwa:C ‘cloud’. It is however very likely that one source of PT *ʰw- is *pw- at a level deeper than PT. Also, note an intriguing distributional fact about this PT voiceless glide. No instances of PT *ʰw- attested so far precede a PT front vowel. 3.7 Summary In this chapter, I have proposed a system of PT simple onsets. The proposed system resembles the conventional reconstruction (Li 1977) in that it is a rich system which symmetrically contrasts voice and voiceless consonants, both among obstruents and sonorants. However, it departs from the conventional reconstruction in major respects. First, the current system of PT onsets has six contrastive places of articulation. I have shown that a series of uvular onsets must be reconstructed for PT. Second, the proposed system of PT simple onsets views the so-called “voicing alternation” as two distinct phenomena—borrowings and reduction of sesquisyllables. While many etyma can be shown to be late borrowings from Chinese, others are viewed as results of reduction of C̬ .C̥ - sequences. Furthermore, this proposal holds that PT did not have contrastive aspiration. Aspirated consonants in modern Tai dialects have been shown to be either loans or to have had complex onsets in PT. PT clusters of the type *Cr- have been shown to be one of the sources that gave rise to modern aspirated onsets. The *Cr- and *C̬ .C̥ sequences will be treated together with other PT complex onsets in the following chapter. 136 CHAPTER 4 PROTO-TAI COMPLEX ONSETS 4.1 Introduction In addition to simple onsets, PT phonology also allowed a rich array of complex onsets. Previous works on PT reconstruction propose a sizeable inventory of clusters in PT. Though differing in details, Li (1977), Haudricourt (1948), Nishida (1954), and Sarawit (1973) hold that a PT cluster consisted of an obstruent or a nasal followed by a liquid or glide, e.g. *kr-, *gr-, *xr-, *ŋr-, *kw-, *ɣw-, *w-, *ŋw-, etc. In contrast, Luo (1997) believes that more complex clusters existed, e.g. *zr-, *zɬ-, *ɟl-45, *ztr-, *ɕkʰr-46, *sb-, etc. Similarly, Liang and Zhang (1996) posit clusters like *zd-, *xpl-, *xkw- etc. However, it is safe to say that these proposals all assume, at least implicitly, that these PT clusters functioned as onsets of monosyllabic words. In contrast to this monosyllabic view is the view that the PT prosodic word could be larger than a simple monosyllable. As established in Chapter 2, the complex onset of a PT prosodic word, according to the current reconstruction, can be either a tautosyllabic or sesquisyllabic cluster. Each monosyllabic word consisted of an obligatory onset and a heavy rime. Similarly, the major syllable of a sesquisyllabic word was composed of an onset and a heavy rime. The minor syllable consisted of only one syllabic consonant or one onset plus an optional syllabic consonant. For convenience, the minor syllable and the onset of the major syllable in a sesquisyllabic word are treated as a single complex onset. A complex onset of a monosyllable is labeled here as a “tautosyllabic cluster” while a complex onset of a sesquisyllable is referred to here as a “sesquisyllabic cluster”. 45 *ǰl- in Luo’s notation 46 *škhr- in Luo’s notation 137 In this chapter, I propose a system of PT complex onsets according to the phonotactics of the PT prosodic word established in Chapter 2. I first present an inventory of tautosyllabic clusters in PT before proceeding to identify their sesquisyllabic counterparts. Then, I discuss the processes involved in the reduction of sesquisyllables to monosyllables. 4.2 Tautosyllabic clusters PT tautosyllabic clusters could consist of up to three consonants. As discussed in §2.3, these consonants strictly obeyed the Sonority Sequencing Principle (Clements 1990). Most consonants could function as the first element in a tautosyllabic cluster but only the most sonorous segments can be in medial positions. Therefore, only liquids and glides were allowed to occupy the second position in a cluster. Table 4-1 presents the inventory of CC- clusters that can be reconstructed for PT. Permissible tautosyllabic clusters in PT consisted of either two or three consonants. Most clusters consist of two consonants. In these two-consonant clusters, only stops can be combined with medial *-r-, and *-l-, e.g. *pr-, *br-, *bl-, *ɓl- etc. In contrast, consonants of any manner of articulation can all be combined with medial *-w- to form clusters, e.g. *pw-, *bw-, *ʰmw-, *ʰlw-, etc. In addition, a few threeconsonant tautosyllabic clusters can be reconstructed, e.g. *krw-, and *klw-. 138 Table 4-1 CC- clusters in PT labial alveolar palatal velar uvular voiceless stops implosives voiced stops -r-l-w-l-r-l-w- fricatives -w- *pr*pl*pw*ɓl*br*bl*bw- *tr*tw- *sw nasals liquid *ʰmw-w- *nw*ʰrw-w*rw- *cr*ɲw- *kr*kl*kw- *gr(*gl-) *xw*ɣw- *ŋw- *qr*qw- *ɢw- Although the second consonant of a cluster can be either *-r-, *-l-, or *-w-, only the labial glide *-w- can occupy the third position in a tautosyllabic cluster. Some three-part tautosyllabic clusters in PT included *krw-, *klw-, *crw-, *trw-, to cite a few. This is predicted because only *-w- combines as a second member of biconsonantal clusters with initial sonorants. It is important that the medial liquids and glides in tautosyllabic clusters did not contrast for voicing. For example, there is no contrast between *k- plus *r-, and *k- plus *ʰr-. Although the medial consonants are transcribed as voiced *-r-, *-l-, and *-w-, these sonorants were not specified for voicing phonologically. As far as the tonal register split is concerned, the first consonant of a cluster determined the series of the tone reflexes in modern dialects. 139 For example, PT *pr- is reflected with tones in the first series while *br- is reflected with tones in the second series. In the following section, Siamese (SI), Lungchow (LC), and Yay (Y) represent SWT, CT, and NT, respectively. 4.2.1 Clusters with medial *-rThere were numerous clusters with medial *-r- in PT but only stops were allowed to occupy the first position of such a cluster. Moreover, voicing and places of articulation both seem to have been an important factor. While all voiceless stops occurred with a medial *-r-, among voiced stops, only the voiced labial and velar ones were allowed to combine with the liquid. PT tautosyllabic clusters with medial *-rincluded *kr-, *tr-, *cr-, *kr-, *qr-, *br- and *gr-. Recall that the *C̥ r- clusters went through aspiration due to the medial *-r- in CT and SWT (see §3.5.2). 4.2.1.1 *prThis PT clusters developed along two paths in different dialects. In SWT, CT dialects, and Saek, it became aspirated to *pʰr- as discussed in §3.5.2. This aspirated cluster simplified to /pʰ-/ in all SWT dialects including Siamese, Lao, Aiton, Black Tai, and White Tai, and in such CT dialects as Shangsi, Western Nung, Guangnan Nung, and Yanshan Nung. Moreover, the medial *-r- also yodicized to /-j-/ in many CT dialects including Lungchow, Leiping, Debao, Chongzuo, and Daxin. In Yongnan and Long’an, PT *pr- is reflected as /pʰl-/. Saek is unique in showing /pʰr-/, attesting the intermediate stage immediately after the aspiration of *C̥ r- clusters. In NT, the cluster did not aspirate and is now reflected as /pl-/, /pj-/, /pɣ-/, or simple /p-/. This cluster corresponds to Li’s *pʰl/r- (Li 1977: 87-89). Table 4-2 gives example of etyma with PT *pr-. 140 Table 4-2 Etyma with PT *pr- Gloss ‘vegetable’ ‘head hair’ ‘lean’ ‘bamboo’ ‘taro’ PT SI LC Y Others *prakD pʰakDS1 pʰjakDS1 piəkDS1 Saek /pʰrakDS1/ *prɤmA pʰomA1 pʰjumA1 piəmA1 Saek /pʰramA1/ *pro:mA pʰɔ:mA1 pʰjo:mA1 pjomA1 Saek /pʰrɔ:mA1/ *prajB pʰajB1 pʰajB1 Saek /pʰrajB1/ *prɯəkD pʰɯəkDL1 pʰɤ:kDL1 pɯəkDL1 Note that in Lungchow the medial *-j- from PT *-r- was lost if the PT rimes contained a high unrounded vowel *ɯ or *i, or a final *-j, e.g. PT *prɯəkDL1 ‘taro’ > *pʰjɯəkDL1 > *pʰɯəkDL1 > Lungchow /pʰɤ:kDL1/. In addition, Yay -j- became absorbed into the rime if followed by /-a-/, e.g. *prakD ‘vegetable’ > pjakDS1 > Yay /piəkDS1/, and *prɤmA ‘head hair’ > pjamA1 > Yay /piəmA1/. 4.2.1.2 *trIn parallel with *pr-, this PT cluster developed along two paths in different dialects. In SWT, NT, and some dialects of CT, *tr- simplified to *ʰr- merging with the original PT voiceless trill (see §3.6.9.1). Siamese, Black Tai, White Tai, Lue, Wuming, Bouyei, Saek, Hechi, Tiandong, Laibin, Shangsi, Leiping, Lungchow, Yay, etc. all show this development. The modern reflexes of the merger include /r-/, /l-/, /ʰl/, /j-/, /ʰj-/, /ð-/, /ɣ-/ and /h-/. In contrast, a number of dialects kept the initial *t- and later aspirated it under the influence of *-r-. Western Nung, Debao, Guangnan Nung, Bao Yen, Wenma, etc. developed along this path. The resulting cluster is now reflected as a plain aspirated /tʰ-/. This cluster corresponds to Li’s *tʰr- (Li 1977: 121125). Table 4-3 gives example of etyma with PT *tr-. 141 Table 4-3 Etyma with PT *tr- Gloss ‘carrying pole’ ‘to see’ ‘stone’ ‘head louse’ ‘loom’ PT *tra:pD *tranD *tri:lA *trawA *trukD SI LC ha:pDL1 ha:pDL1 henA1 -v hanA1 hinA1 hinA1 hawA1 hawA1 hu:kDL1 hukDL1 Y ra:pDL1 ranA1 rinA1 rawA1 rokDL1 Others Debao /tʰa:pDL1/ Debao /tʰanA1/ Debao /tʰənA1/ Debao /tʰawA1/ Debao /tʰo:kDL1/ 4.2.1.3 *crThis PT cluster developed differently from the other *C̥ r- clusters. In SWT and most CT varieties, this cluster became aspirated to *cʰr- in parallel with *pr-, *tr-, and *kr-. Ferlus (1990)’s *cʰr- reflects this intermediate step. This cluster later became *tʰr- and is now reflected as /tʰ-/. Siamese, Lao, Black Tai, Lue, Lungming, Lungchow, Jingxi, Ningming, Wenma, and Cao Bang are examples of varieties in this group. The only NT dialect that went through this development is Saek. In many NT dialects including Liujiang, Rong’an, Nandan, Lianshan, etc. *cr- dissimilated to *krafter the original *kr- had become *ʰr- (see §4.2.1.4). In these dialects, the cluster is now reflected /kj-/. Another development is found in some CT and NT varieties including Shangsi, Long’an, Wuming, Pinguo, Qiubei, Guigang, etc. In this group of dialects, the complex onset *cr- simplified and merged with PT *c- (see §3.6.1.3). It is now reflected as /c-/, /ts-/, /s-/, /ɕ-/. This cluster is attested only in a few examples in Table 4-4 and corresponds to Li’s *tʰl- (1977: 119-120). 142 Table 4-4 Etyma with PT *cr- Gloss ‘to wait’ ‘bag’ ‘senior’ PT *cra:C *croŋA *crawC SI tʰa:C1 tʰuŋA1 tʰawC1 LC tʰa:C1 tʰuŋA1 tʰawC1 Y sa:C1 sɔŋC1 Others Guigang /ca:C1/ Note that ‘old’ is assigned here only tentatively because NT evidence is lacking. 4.2.1.4 *krThis PT cluster developed along two paths in different dialects. In SWT, NT, and many CT dialects, it was simplified to *ʰr- merging with the original PT voiceless trill (see §3.6.9.1). Siamese, Black Tai, White Tai, Lue, Yay, Wuming, Bouyei, Saek, Hechi, Tiandong, Laibin, Shangsi, Leiping, Lungming, Lungchow, etc. all show this development. The modern reflexes of the merger include /r-/, /l-/, /ʰl-/, /j-/, /ʰj-/, /ð-/, /ɣ-/ and /h-/. Another path is taken by a number of CT dialects including Western Nung, Debao, Guangnan Nung, Bao Yen, and Wenma, to cite a few. In these dialects, *kr- did not simplify but later went through aspiration due to the medial *-r- (see §3.5.2), which gave *kʰr-. The resulting aspirated cluster is now reflected as /kʰj-/ or / /cʰ-/. This cluster corresponds to *xr- in Li (Li 1977: 233-235)’s reconstruction. 143 Table 4-5 Etyma with PT *kr- Gloss ‘to seek’ ‘gust (of rain)’ ‘six’ ‘to laugh’ ‘fish net’ PT *kra:A *kra:B *krokD *krɯəwA *kre:A SI ha:A1 ha:B1 hokDS1 huəA1 hɛ:A1 LC ha:A1 ha:B1 hokDS1 hu:A1 he:A1 Y ra:A1 ra:B1 rokDS1 riəwA1 reA1 Others Debao /kʰja:A1/ Debao /kʰja:B1/ Debao /kʰjokDS1/ Debao /kʰuA1/ Debao /kʰeA1/ 4.2.1.5 *qrThis cluster went through two main lines of development. In SWT as well as many CT and NT dialects, *qr- changed to *kr-, but did not merged with the original *kr- (see §4.2.1.4). The intermediate cluster is still retained as /kr-/ and /kɣ-/ in Guigang and Laibin respectively but became /kl-/ in Hengxian. In a number of other dialects, the medial *-r- yodicized to *-j-, which further palatalized the *k- to /c-/. Thus, while Pingguo, Long’an, Longsheng, Du’an, etc. have /kj-/ for this PT cluster, Yay, Tiandong, Tianlin, Lingyue, etc. show /c-/. In SWT and most CT dialects, the intermediate cluster *kr- became aspirated due to the influence of *-r-. Now it is reflected as /kʰj-/ in Leiping, Lungchow, Debao, and Ningming; as /cʰ-/ in White Thai, Western Nung, Guangnan Nung, and Yanshan Nung; as /s-/ in Black Tai and Daxin; and as /kʰ-/ in Siamese, Lao, Lue, etc. The NT language Saek also went through aspiration and now reflects this cluster as /tʰr-/ from an intermediate *kʰr-. Another line of development is a full merger with *ʰr- as illustrated by a number of CT dialects which simplified *qr- to *ʰr- in parallel to PT *kr- (see §4.2.1.4). The reflex of PT *qr- in Lungming, Fusui, Qinzhou and Shangsi is a simple /l-/ while its reflex in Yongnan and Long’an is a /ʰl-/. Etyma that have this PT cluster 144 are listed under *kʰl- in Li’s reconstruction (Li 1977: 226-228). Only the two examples given in Table 4-6 have been attested for this PT cluster. Also see *k.r- in §4.4.4 for a similar development. Table 4-6 Etyma with PT *qr- Gloss ‘egg’ ‘to sift’ PT *qrajA *qrɤŋA SI kʰajB1 kʰɯŋA1 LC kʰjajB1 kʰjɤŋB1 Y cajB1 Others Lungming /lajB1/ Lungming /lɤŋB1/ 4.2.1.6 *brThe reflex of this cluster in modern dialects depends largely on whether PT voiced stops are reflected as plain or aspirated stops (see §3.6.3), and what the general reflex of medial *-r- is (see §4.2.1). In Lungchow, PT *b- was devoiced to voiceless *p-, and PT *-r- yodicized to /-j-/. Therefore, the modern reflex of *br- in these varieties is /pj-/, i.e. *br- > *bj-, > /pj-/. Similarly, it is reflected in Lao as a simple /pʰ/ because *b- regularly became /pʰ-/ and medial *-r- was regularly dropped, i.e. *br- > *pʰr- > /pʰ-/. Reflexes of this cluster found in modern dialects include /bj-/, /p-/, /pl-/, /pr-/, /pɣ-/, /pj-/, /pʰr-/, /pʰj-/, and /pʰ-/. This cluster is attested by only a few examples in Table 4-7 but they are attested in all groups of dialects and show regular correspondences. Li (1977: 89-91) also reconstructs *br- for these etyma. 145 Table 4-7 Etyma with PT *br- Gloss ‘to separate’ ‘to deceive’ ‘deficient’ PT *bra:kD *bra:ŋA *bro:ŋB SI pʰra:kDL2 pʰra:ŋA2 pʰrɔ:ŋB2 LC pja:kDL2 pja:ŋA2 pjo:ŋB2 Y pja:kDL2 pjɔ:ŋB2 Others Cao Bang /b̤ ja:kDL2/ Cao Bang /b̤ ja:ŋA2/ Cao Bang /b̤ jɔŋB2/ 4.2.1.7 *grLike *br-, the reflex of this cluster in modern dialects depends largely on what the reflexes of *g- (see §3.6.3.4) and *-r- (see §3.6.8.1) are. For example, PT is reflected as *kj- in Debao because *g- and *-r- are reflected as *k- and *-jrespectively. Furthermore, some dialects palatalized the velar stop due to the influence of /-j-/, e.g. Leiping *gr- > *kʰj- > /cʰj-/. Therefore, reflexes of this cluster found in modern dialects include /kj-/, /k-/, /kl-/, /kr-/, /kɣ-/, /kj-/, /kʰr-/, /kʰj-/, /kʰ-/, /c-/ and /cʰ-/. A special development is found in Lungming, where *gr- is reflected as /l-/. This PT cluster is attested by only few examples in Table 4-8. Li (1977: 230-232) also reconstructs *gr- for these etyma. Table 4-8 Etyma with PT *gr- Gloss ‘mortar’ ‘to moan’ ‘litter (of young animal)’ ‘half’ ‘indigo’ PT *grokD *gra:ŋA *gro:kD *grɯŋB *gra:mA SI kʰrokDS2 kʰra:ŋA2 kʰrɔ:kDL2 kʰrɯŋB2 kʰra:mA2 LC Y Others kjukDS2 cokDS2 Guigang /kra:ŋA2/ kjo:kDL2 cokDL2 Wuming /klaŋB2/ kja:mA2 sa:mA2 -i 146 4.2.2 Clusters with medial *-lThere were a number of clusters with medial *-l- in PT but only stops were allowed to occupy the first position of such a cluster. Like cluster with *-r-, places of articulation seems to have been an important factor. PT tautosyllabic clusters with medial *-l- included only *pl-, *kr-, *bl-, *gl-, and *ɓl-. Recall that the medial *-l- did not aspirate the preceding stop (see §3.5.2). Therefore, in many dialects the contrast between *C̥ l- and *C̥ r- is still preserved, e.g. Lungchow /pja:A1/ from PT *pla:A ‘fish’ vs. /pʰja:A1/ from PT *pra:A ‘stone mountain’. 4.2.2.1 *plThis PT cluster is reflected as /p-/, /pl-/, /pj-/, /pr-/ or /pɣ-/ in modern dialects. While dialects like Siamese, Saek, Yongnan, and Hengxian still preserve the cluster as /pl-/, Lao, White Tai, Black Tai, Western Nung, Qiubei, and Du’an simplified it to /p-/. In a great number of varieties including Lungchow, Lungming, Bao Yen, Chongzuo, Jingxi, Tianlian, Rong’an, Nandan, etc., the medial *-l- yodicized to /-j-/, resulting in /pj-/. In Laibin and Guigang, *pl- completely merged with *pr- and is now reflected as /pɣ-/ and /pr-/ respectively. Li (Li 1977: 84-86) also reconstructs *pl- for this set of data. Table 4-9 presents examples of etyma with PT *pl-. 147 Table 4-9 Etyma with PT *pl- Gloss ‘fish’ ‘tip, end’ ‘to take down’ ‘to awaken’ ‘aquatic leech’ PT *pla:A *pla:jA *ploŋA *plokD *pli:ŋA SI pla:A1 pla:jA1 ploŋA1 plukDS1 pliŋA1 LC pja:A1 pja:jA1 pjuŋA1 pjukDS1 piŋA1 Y pja:A1 pja:jA1 pjoŋA1 pjokDS1 piŋA1 Others The Lungchow and Yay forms for ‘aquatic leech’ do not have medial /-j-/ because it was absorbed by the high front vowel /i/. 4.2.2.2 *klThis PT cluster is reflected as /k-/, /kl-/, /kj-/, /kr-/, /kɣ-/ or /c-/ in modern dialects. While dialects like Siamese, Yongnan, and Hengxian still preserve the cluster as /kl-/, Lao, White Tai, Black Tai, and Yongbei have simplified it to /k-/. Saek also still preserves the medial lateral but has changed the cluster to /tl-/. In a great number of varieties including Lungchow, Lungming, Bao Yen, Chongzuo, Jingxi, Rong’an, Nandan, etc., the medial *-l- vocalized to /-j-/, resulting in /kj-/. In Leiping, Guangnan Nung, Tiandong, and Qiubei the cluster further palatalized to /c-/. Yay, Laibin and Guigang, to cite a few, completely merged medial *-l- and *-r- so that *kl- is now reflected /kɣ-/ and /kr-/ respectively. Li (Li 1977: 220-225) also reconstructs *kl- for this set of data. Table 4-10 presents examples of etyma with PT *kl-. 148 Table 4-10 Etyma with PT *kl- Gloss ‘middle’ ‘banana’ ‘rice seedling’ ‘drum’ ‘last (year)’ PT *kla:ŋA *kluəjC *kla:C *klo:ŋA *kla:jA SI kla:ŋA1 kluəjC1 kla:C1 klɔ:ŋA1 kla:jA1 LC kja:ŋA1 ku:jC1 kja:C1 kjo:ŋA1 kja:jA1 Y ca:ŋA1 cuəjC1 ca:C1 coŋA1 ca:jA1 Others 4.2.2.3 *ɓlThere are three main lines of development for *ɓl-. The first one is simplification to *ɓ- in Black Tai, White Tai, and Shangsi. In these dialects, it is now reflected as plain /b-/. The second line of development is coalescence to *ɗ-, merging with the original PT *ɗ- (see §3.6.2.2). This development is found in Siamese, Lao, Western Nung, Debao, Yay, Rong’an, Hechi, etc. In these dialects, *ɓl- is now reflected as /d-/, /n-/, /l-/ or /ɗ-/. The third line of development is retention of *ɓl-. Dialects in this group reflect the PT cluster as /bj-/, /mj-/, or /bl-/. In most dialects, the medial *-l- later yodicizied to /-j-/, creating /bj-/. Examples of these dialects include Lungming, Lungchow, and Chongzuo. Saek is unique in still retaining the medial -l-. Li (1977: 91-94) includes this set of data under *ˀbl/r-. 149 Table 4-11 Etyma with PT *ɓl- Gloss ‘moon’ ‘flower’ ‘to weed’ ‘gall bladder, bile’ ‘to snap’ PT *ɓlɯənA *ɓlo:kD *ɓla:jA *ɓli:A *ɓli:tD SI dɯənA1 dɔ:kDL1 da:jA1 di:A1 di:tDL1 LC Y bɤ:nA1 dɯənA1 bjo:kDL1 dokDL1 bja:jA1 da:jA1 di:A1 -i di:A1 Others Saek /bliənA1/ Saek /blɔ:kDL1/ Saek /bli:A1/ Saek /bli:tDL1/ Note an unexplained split in Shan dialects, in which PT *ɓl- is reflected as either /m-/ (< *ɓ-) or l- (< *ɗ-) depending on the etyma, cf. Southern Shan /mɔkDL1/ from *ɓlo:kDL1 ‘flower’ vs. /lɤnA1/ from *ɓlɯənA1 ‘moon’. It is possible that the split represent a distinction in PT between *ɓr- as in ‘flower’ and *ɓl- as in ‘moon’. However, Saek, which normally preserves the distinction between medial *-l- and *-r-, only shows /-l-/. Therefore, I tentatively view the split reflex in Shan as an innovation47. 4.2.2.4 *glThis cluster is tentative because only two etyma are found. The first one is ‘nearly alike’, which is reflected as /kʰlapDS2/ in Siamese compound /kʰlapDS2 kʰla:jC2/ ‘vaguely resembling’ (Li 1977: 230), and as /capDS2/ ‘just right’ in Po-ai, and /kjapDS2/ ‘just right’ in Dioi. Another etyma is ‘to lasso’, whose modern forms are / kʰlɔ:ŋC2/, 47 Li (1977: 91) says that Shan, Lao and Lue all show the split but disagree in their reflexes. In contrast, I believe that the regular reflex of *ɓl- in Lao and Lue is /d-/. The /b-/ reflected in some etyma is a special development. For example, Lao /bi:A1/ for ‘gall bladder’ is probably contaminated by /bɯ:A1/ ‘umbilical cord’. 150 /kjo:ŋC2/, and /kløŋC2/ in Siamese, Lungchow, and Wuming respectively. This cluster is also reconstructed as *gl- in Li’s system (1977: 229-230). 4.2.3 Clusters with medial *-wTautosyllabic clusters abound in PT as medial *-w- can be combined with many consonants including stops, fricatives, nasals, and liquids. However, it is important to note that some clusters are more robustly supported than others. While such clusters as *kw-, and *ɣw- are attested by many etyma, *ʰmw- is supported by only one example. 4.2.3.1 Velar + *-wThe most common complex onsets with medial *-w- in PT had velar consonants as the first part of the cluster. These include *kw-, *xw-, *ɣw-, and *ŋw-. Interestingly, *gw- is not attested. It is possible that it merged with *ɣw- early on. While modern dialects usually preserve *kw- and *xw-, they have simplified one or more of the clusters *ɣw- and *ŋw-. For example, Siamese retains the medial *-w- in *kw-, *xw-, and *ɣw- but simplified *ŋw- to a simple /w-/. In contrast, Lungchow and Yay keep *kw- as /kw-/, but simplified the other three to a simple *w-, which is now reflected as /v-/. In Li’s system, these four clusters are also reconstructed as *kw- (Li 1977: 236-237), *xw- (Li 1977: 240-241), *ɣw- (Li 1977: 241-242), and *ŋw- (Li 1977: 239-240). These clusters are exemplified in Table 4-12. 151 Table 4-12 Etyma with PT *kw-, *xw-, *ɣw- and *ŋw- respectively Gloss ‘to sweep’ ‘to cross over’ ‘deer’ ‘axe’ ‘putrid’ ‘water buffalo’ ‘smoke’ ‘penis’ ‘daytime’ ‘dumb’ PT SI LC Y Others *kwa:tD kwa:tDL1 kwa:tDL1 Lungming /kwa:tDL1/ *kwa:B kwa:B1 kwa:B1 kwa:B1 *kwa:ŋA kwa:ŋA1 kwa:ŋA1 vɯəŋA2 *xwa:nA kʰwa:nA1 kʰwa:nA1 va:nA1 White Tai /xwa:nA1/ *xwi:wA kʰiəwA1 kʰi:wA1 Bao Yen /xwi:wA1/ *ɣwa:jA kʰwa:jA2 va:jA2 va:jA2 White Tai /xwa:jA2/ *ɣwanA kʰwanA2 vanA2 White Tai /xɔnA2/ *ɣwajA kʰuəjA1 White Tai /xwajA2/ *ŋwanA wanA2 vanA2 vanA2 Yongnan /ŋonA2/ *ŋwamC vamC2 Yongnan /ŋamA2/ The Yay form for ‘deer’ shows a special development. The cluster *kw- must have been simplified to *w- due to the following vowel *ɯə48, i.e. PT *kwɯəŋA > *wɯəŋA > vɯəŋA2. Moreover, this form shows tone A2, indicating that *kw had simplified to *w- before the binary register split. 4.2.3.2 Uvular + *-wPT clusters with a uvular obstruent followed by *-w- are attested by a few examples in Table 4-13. *qw- completely merged with *xw- in most dialects but in Phu Thai and Phuan dialects they remain distinct. For example, the Kapong dialect of Phu Thai has /kʰw-/ for *xw-, but /w-/ for *qw- via an intermediate stage *ʰw-. The voiced cluster *ɢw- is found only in *ɢwɯn ‘human’ The medial -w- simply dropped 48 [k-], [w-] and [ɯ] are all produced at the velum. 152 out in CT and SWT dialects but left traces in the vowel in NT. In all SWT, including Siamese, White Tai, Lue, Lao, etc., the vowel *ɯ lowered to *ɤ regularly before a non-velar nasal coda. This *ɤ is now reflected as /o/ in SWT (see §5.6.2.3). In many NT varieties including Yay, Saek, Tianlin, Lingyue, Donglan, etc., it had a rounding effect on the following vowel so that now the most common vocalic nucleus in this etymon is /u/. However, in a few dialects, including Pingguo and Nandan, the *ɢwwas simplified *w-, e.g. Pingguo Pingguo /wɯnA2/. While the etyma reconstructed with PT cluster *qw- in the current system are included in Li (1977: 240-241) under *xw- (Li 1977: 240-241), *ɢw- is listed under and *ɣw- (Li 1977: 214-219, 272). Table 4-13 shows etyma with PT *qw-, and *ɢw. Table 4-13 Etyma with PT *qw-, and *ɢw- respectively Gloss ‘to hang’ ‘face down’ ‘spirit’ ‘human, person’ PT SI LC *qwe:nA kʰwɛ:nA1 *qwamC kʰwamC1 kʰumC1 *qwanA kʰwanA1 kʰwanA1 *ɢwɯnA kʰonA2 kʰɤnA2 Y ve:nA1 hamC1 vanA1 hunA2 Others Kapong /wɛ:nA1/ Kapong /wamC1/ Kapong /wanA1/ Notice that the *qw- in ‘face down’ and the *ɢw- in ‘human’ became simplified early on in Yay, i.e. PT *qwamC > *hwamC > /hamC1/ ‘face down’, and PT *ɢwɯnA > *hwɯnA2 > /hunA2/ ‘human’. In Yay, /h-/ is the regular reflex of *qw- and *ɢw-. 4.2.3.3 Labial + *-wFour labial clusters with *-w- have been identified for PT: *bw-, *pw-, *mw- and *ʰmw-. The cluster *mw- is amply attested by reflexes in modern dialects. 153 Normally, PT *m- is reflected as /m-/ with tones in the second series in all dialects. However, there is a set of etyma which are reflected as /m-/ in SWT and CT but as /f-/ in most NT dialects. In other words, the SWT and CT reflex points to earlier *m- (see §3.6.6.1) but the NT reflex points to *w- (see §3.6.8.3). Following Li (1977: 73-74), I reconstruct *mw- for this set of etyma49. Table 4-14 Etyma with PT *mw- Gloss ‘clf. for tools’ ‘hand’50 ‘day’ ‘drunk’ ‘sprirt’ PT *mwa:kD *mwɯ:A *mwɯ:C *mwi:A *mwa:ŋA SI mɯ:A2 mɯ:C2 LC ma:kDL2 mɯ:A2 mɯ:C2 Y fa:kDL2 fɯŋA2 fɯ:C2 fiA2 fa:ŋA2 Others W. Nung /mi:A2/ Chongzuo /ma:ŋA2/ While evidence for *mw- is the most robust, *bw-, *pw-, and *ʰmw are supported by only a few examples. These etyma are shown in Table 4-15. In Li’s reconstruction, they are reconstructed as *b- (Li 1977: 66), *pɰ- (Li 1977, 278-279), and *ʰmw- (Li 1977: 75-76, 263) respectively. 49 It is also possible that ultimately this set of etyma had *r.m- in PT. Although Proto-Kra (Ostapirat 2000) *r-maŋA ‘spirit’ and *r-meA ‘drunk’ suggest that these etyma also go back to PT *r.m-, Tai internal evidence alone does not indicate sesquisyllabic onsets. 50 The final /-ŋ/ found in many NT dialects is a sporadic development, possibly PT *r.mɯ:A > *r.mɯ:ŋA > Yay /fɯŋA2/. Compare with PT ‘2s pronoun’ in Appendix B. 154 Table 4-15 Etyma with PT *bw-, *pw-, and *ʰmw- Gloss ‘fat’ ‘lung’ ‘bear’ ‘pubic hair’ PT *bwi:A *pwɤtD *ʰmwɯjA *ʰmwu:j A SI pʰi:A2 pɔ:tDL1 mi:A1 mɔ:jA1 LC pi:A2 pɤtDS1 mi:A1 Y Others pi:A2 Huanjiang /pwi:A2/ putDS1 Nandan /pɯətDS1/ mɯəjA1 Huanjiang /mu:jA1/ Po-ai /mi:A1/ For PT *bwi:A ‘fat’, the medial *-w- is attested in Huanjiang as well as Rong’an and Longsheng. Further support comes from its Chinese source 肥 féi, whose reconstructed MC form is bjwei. For PT *D pʰjwɐiC ‘lung’, the *-w- is most likely preserved as /ɯ/ in Nandan. Moreover, the Late Han pronunciation of its Chinese source 肺 fèi is pʰuɑs, which lends further support to the reconstruction of medial *-w-. As for *ʰmwɯjA ‘bear’ and *ʰmwu:jA ‘pubic hair’, the medial *-w- is not attested as a separate segment but traces of its rounding are still found in the rimes. 4.2.3.4 Coronal + *-wMany coronal consonants in PT can be combined with medial *-w- but the clusters are usually simplified to simple onsets. At minimum, the following complex onsets must be reconstructed for PT: *tw-, *nw-, *rw-, *ʰrw-, and *ʰlw-. These clusters are reconstructed as simple onsets *t-, *n-, *r-, *ʰr-, and *ʰl- in Li. The most common coronal cluster with medial *-w- is perhaps *rw- shown in Table 4-17. 155 Table 4-16 Etyma with PT *rw- Gloss ‘to pour’ ‘to vomit’ ‘gutter, trough’ ‘chicken louse’ ‘dry field’ PT *rwa:tD *rwɯəkD *rwɯəŋA *rwɤjA *rwɤjB SI ra:tDL2 ra:ŋA2 rajA2 rajB2 LC ɬa:kDL2 ɬa:ŋA2 ɬajA2 ɬajB2 Y Others Kapong /hwa:tDL2/ ruəkDL2 Bouyei(4) /zvɯəDL2/ ruəŋA2 riA2 Yishan /ʰjwiA2/ riB2 Yishan /ʰjwiB2/ For *rwɤjA ‘chicken louse’ and *rwɤjB ‘dry field’, the *-w- is attested in only a few varieties. In Yishan /ʰjwiA2/ and /ʰjwiB2/, PT *r- is reflected as /ʰj-/ and the *-w- is still preserved. For *rwɯəkD ‘to vomit’, and *rwɯəŋA ‘gutter, trough’, the medial glide is generally lost but its rounding is still preserved on the diphthong in Yay /ruəkDL2/, and /ruəŋA2/. Similarly, the medial /-v-/ in Bouyei(4) for ‘to vomit’. Moreover, /-w-/ in Kapong /rwa:tDL2/ ‘to pour over’ and the rounded vowel in Po-ai /lu:tDL2/ attest the PT medial *-w-. In addition to *rw-, a few other coronal clusters existed in PT. Examples of these clusters are given in Table 4-17. 156 Table 4-17 Etyma with PT *tw-, *nw-, *ʰrw-, and *ʰlw- respectively Gloss ‘to crawl over’ ‘jackal’ ‘way’ ‘mushroom’ ‘comb’ ‘to flow’ PT *twajB *nwajA *ʰrwɤnA *ʰrwetD *ʰrwɯ:jA *ʰlwajA SI tajB1 najA2 honA1 hetDS1 wi:A1 lajA1 LC tajB1 vitDS1 vi:A1 lajA1 Y najA1 ranA1 ratDS1 rojA1 lajA1 Others Bao Yen /twajB1/ Bao Yen /nwajA1/ Huanjiang /wanA1/ Yishan /ʰjotDS1/ Nandan /ruəjA1/ Bao Yen /lwajB1/ For *twajB ‘to crawl’, *nwajA ‘jackal’, and *ʰlwajA ‘to flow’, the medial *-wis clearly attested in a number of dialects, cf. Bao Yen /twajB1/, /nwajA1/, /lwajB1/ respectively. For *ʰrwɤnA ‘way’, the medial *-w- is preserved as initial /w-/ as in Huanjiang /wanA1/, or as rounding of the vowel as in Pingguo /ronA1/. Similarly, the medial *-w- in *ʰrwetD ‘mushroom’ is preserved as /v-/ as in Lungchow /vitDS1/ or as vowel rounding as in Yishan /ʰjotDS1/. Similarly, the medial *-w- in *ʰrwɯ:jA ‘comb’ is preserved as initial /w-/ in Siamese /wi:A1/, /v-/ in Lungchow /vi:A1/, and as rounding in most NT dialects as in Yay /rojA1/ and Nandan /ruəjA1/. 4.2.3.5 Three-consonant clusters *CCwPT allowed three consonants to form tautosyllabic clusters under the constraints discussed in §4.2. At a minimum, *krw-, *klw, *qrw- *trw- and *crwmust be reconstructed at the PT level. Examples of these clusters are given in Table 4-18. Li (Li 1977: 120, 227-228) reconstructs two-consonant clusters for these etyma. 157 Table 4-18 Etyma with PT *krw-, *klw-, *qrw-, *trw-, and *crw- respectively Gloss PT SI LC Y Others ‘spider’ *krwa:wA kʰja:wA1 ca:wA1 Wuming /klwa:wA1/ ‘small frog’ *krweA kweA1 Wuming /klweA1/ ‘salt’ *klwɯəA klɯəA1 kɯ:A1 kuəA1 Rong’an /kwɤA1/ ‘mountain stream’ *qrwɤjC huəjC1 vu:jC1 viC1 Longsheng /jwiC1/ ‘to carry on the back’ *trwa:mA ha:mA1 ha:mA1 ra:mA1 Longsheng /jwa:mA1/ ‘plough, to plough’ *crwajA tʰajA1 tʰajA1 sajA1 Longsheng /kjwaiA1/ For *krwa:wA ‘spider’ and *krwe:A ‘small frog’, the clusters usually simplified to either *kr- or *kw- but a few dialects including Wuming still retain all three consonants, cf. /klwa:wA1/ and /klweA1/. For *klwɯəA ‘salt’, not only did the *-wleave rounding on the rime in dialects like Yay, but the Rong’an form also retains the medial. For *qrwɤjC ‘mountain stream’, the medial is reflected both in the /u/ in Siamese /huəjC1/ and the /v-/ in Lungchow /vu:jC1/ and Yay /viC1/. For *trwam and *crwaj, the medial *-w- is lost in most dialects, but is preserved in some dialects including Longsheng, cf. /jwa:mA1/ and /kjwaiA1/. Note that both initial *r- and medial *-r- became /j/ in Longsheng. 4.3 Speculative *-wIn addition to the complex onsets with medial *-w- above, I also speculate that medial *-w- is responsible for some vowel correspondences. This speculative *-w- is posited in both monosyllabic and sesquisyllabic PT etyma. There are altogether two paradigms of vowel correspondences for which I posit a *-w- despite a lack of unequivocal evidence. 158 The first paradigm involves etyma whose SWT and CT reflexes points to an earlier long mid vowel but whose NT reflexes point to earlier short high vowels. For example, Siamese /tʰɔ:ŋC2/ suggests earlier *do:ŋC but Yay /tuŋC2/ suggests earlier *duŋC. In this case, I speculate that a medial *-w- had a lowering effect on following *u:, resulting in a long mid vowel *o: in Siamese. In Yay, the *-w- simply dropped out, leaving *u: as the reflex. This situation is in contrast with the PT *o:, which is reflected as /ɔ:/ in Siamese, /o:/ in Lungchow, and /o/ in Yay. It also differs from PT u:, which is reflected as /u:/ in Siamsese, /u/ in Lungchow, and /u/ in Po-ai. Table 4-19 presents some etyma for which I propose *wu:. Li reconstructs *uɔ (Li 1977: 278), which contrasts with his *wɔ (Li 1977: 277-278). Table 4-19 Etyma with PT *wu: Gloss ‘stomach, belly’ ‘young, soft’ ‘to dye’ ‘spool’ PT SI *dwu:ŋC tʰɔ:ŋC2 *ʔwu:nB ʔɔ:nB2 *ɲwu:mC ɲɔ:mC2 *lwu:tD lɔ:tDL1 LC to:ŋC2 ʔo:nB2 jo:mC2 lo:tDL1 Y tuŋC2 ʔunB2 ɲumC2 Others Guigang /ʔwunB2/ Po-ai /lutDS1/ Note that the NT reflex for ‘spool’ point to earlier short vowels rather than long vowels because of the regular shortening of PT long *u: in closed syllables (see §5.6.1.3). Although reconstructing *-w- for this set of data is very speculative, there seems to be some concrete evidence for two of the etyma. The first one is ‘soft’ whose reflex in the NT dialect of Guigang shows a medial /-w-/. The second one is ‘to dye’ whose Vietnamese counterpart nhuộm ‘to dye’ has a diphthong /uə/ as vocalic nucleus. 159 The speculative *-w- is also posited for the data involving front vowels in Table 4-20, for which Li (Li 1977: 274-275) reconstructs *iɛ and *jɛ. Table 4-20 Etyma with PT *-wi:- Gloss ‘sunshine’ ‘late meal’ ‘red’ ‘wasp’ PT *C̥ .dwi:tD * C̥ .lwi:ŋA *C̥ .dwi:ŋA *b.twi:lA SI dɛ:tDL1 lɛ:ŋA2 dɛ:ŋA1 tɛ:nA1 LC de:tDL1 le:ŋA2 de:ŋA1 pʰe:nA1 Y ditDS1 riŋA2 diŋA1 tinA2 Others PT *e: is regularly reflected as /ɛ:/ in Siamese, /e:/ in Lungchow, and /e/ in Yay, while PT *i: is reflected as /i:/, /i/, and /i/ in the three languages respectively. The cases above, in contrast, show a different correspondence: Siamese /ɛ:/, Lungchow /e:/ but Yay /i/. I speculate that in these cases a medial *-w- lowered the following *i: to a long mid vowel *e: in CT and SWT before dropping out. In NT, the *-w- simply was lost, leaving *i: as the reflex. For example, Siamese /dɛ:ŋA1/ came from earlier *ɗe:ŋA, whose long *e: in turn resulted from lowering of PT *i: in *C̥ .dwi:ŋA. In contrast, Yay /diŋA1/ came from earlier *ɗiŋA, whose *i was a regular reflex of PT *i: after the speculated medial *-w-. Note that the Yay reflex for ‘sunshine’ has DS1 rather than DL1 because of the regular shortening of PT long *i: in closed syllables (see §5.6.1.1). So far, I have not been able to identify potential evidence for the medial *-w-. Therefore, the reconstruction of these cases remain speculative. Note that it is not possible, however, to posit *-j- instead of *-w- because this paradigm would be the only set of data that had the medial *-j-. 160 The second paradigm involves etyma whose SWT and CT reflexes point to an earlier long mid vowel but whose NT reflexes point to an earlier diphthong. For example, Siamese /nɔ:ŋC2/ suggests earlier *no:ŋC but Yay /nuəŋC2/ suggests earlier *nuəŋC. In this case, I speculate that the PT form of this etymon was *nwo:ŋC. In this scenario, the medial *-w- was lost in front of the long *o: yielding a simple long *o: in SWT and NT. This *o: is now reflected as /ɔ:/ in Siamese. In NT, the *-w- coalesced with the nucleus *o: to become *uə, which is still preserved in Yay. In addition to this *o:~*uə correspondence, the paradigm also includes a front vowel *e:~*iə set and a back unrounded *a:~*ɯə set. Therefore, for this paradigm, I tentatively propose PT *we:, *wɤ: and *wo: for Li’s *jɛ (1977: 274-275), *wa (1977: 276) and *wɔ (1977: 277-278) respectively. Table 4-21 presents etyma belonging to this paradigm. Table 4-21 Etyma with PT *we:, *wɤ: and *wo: Gloss ‘cucumber’ ‘tail’ ‘downstairs’ ‘crossbow’ ‘to disappear’ ‘bad’ ‘younger sibling’ ‘elbow’ PT SI *p.rwe:ŋA tɛ:ŋA1 *ʰrwɤ:ŋA ha:ŋA1 *lwɤ:ŋA la:ŋB2 *ʰnwɤ:C na:C1 *ʰrwɤ:jA ha:jA1 *rwɤ:jC ra:jC2 *nwo:ŋC nɔ:ŋC2 *C̥ .swo:kD sɔ:kDL1 LC pʰe:ŋA1 ha:ŋA1 la:ŋB2 ɬa:jC2 no:ŋC2 Y tiəŋA1 rɯəŋA1 nɯəC1 rɯəjA1 rɯəjC2 nuəŋC2 suəkDL1 Others Po-ai /lɯ:ŋB2/ W. Nung /na:C1/ Lungming /ha:jC1/ Lungming /so:kDL1/ 161 The only potential support for the speculated *-w- in this paradigm comes from the MC form for 弩 nǔ (< MC nuoB, LH nɑB, OC *nâʔ) ‘crossbow’. Although the u in the MC forms may be viewed as concrete evidence for a medial *-w- in Proto-Tai *nwɤ:C, the chronology is rather problematic. PT most likely predates Late Han Chinese at which stage the rime of 弩 was still simple ɑ. Therefore, the reconstruction of PT *-w- for this set of data must be taken as tentative. 4.4 Sesquisyllabic clusters In §2.3, I established that a PT prosodic word could be either monosyllabic or sesquisyllabic. I argued that this sesquisyllabic view of PT phonology best accounts for the wide range of sound correspondences observed among the daughter languages. According to the proposal, a sequence of consonants that makes up the minor syllable plus the major syllable onset, or a sesquisyllabic cluster, can have up to two consonants in the minor syllable, and up to three consonants in the major syllable. In this section, I identify sesquisyllabic clusters that can be reconstructed for PT before proceeding to discuss the process through which sesquisyllables consisting of these complex onsets reduced to monosyllables in modern languages. Note that some clusters posited here are more speculative than others. Many clusters are attested by only one or two etyma, and must therefore be taken as tentative. 4.4.1 Voiceless stop + voiceless stop Clusters that consist of two voicless obstruents are perhaps the most common type of sesquisyllabic clusters in PT. In many of such *C̥ .C̥ - clusters, both the minor syllable and the major syllable onsets are voiceless stops, e.g. *p.t- and *k.t-. When these complex onsets were reduced to fit the monosyllabic template, the resulting onsets stayed voiceless. In all modern languages, etyma with this type of PT cluster 162 have tones in the first series. The most robust set is PT *p.t- given Table 4-22. Li (1977: 118-119) posits a tautosyllabic *tr- for this set of data. Table 4-22 Etyma with PT *p.t- Gloss ‘eye’ ‘to die’ ‘grasshopper’ PT *p.ta:A *p.ta:jA *p.takD SI ta:A1 ta:jA1 takDS1 LC ha:A1 ha:jA1 Y ta:A1 ta:jA1 takDS1 Others Saek /pra:A1/ Saek /pra:jA1/ Lungming /tʰak/ Following Ferlus (1990), PT *p.t- simplified to /t-/ in Siamese and Yay but went through weakening of the non-initial *.t- in Saek and Lungchow. In Lungchow, original *p.t- became *p.r- after the non-initial *.t- became weakened to *.r-. The *pwas then lost leaving only a voiceless *ʰr-, which is now reflected as /h-/. In Saek, the original *p.t- became *p.r- after weakening but, rather than dropping the initial *p-, the sesquisyllabic cluster became a true tautosyllabic cluster /pr-/. The second set of etyma that go back to a sesquisyllabic cluster consisting of two voiceless stops is presented in Table 4-23. Following Ferlus (1990), I reconstruct PT *k.t- for these etyma, for which Li (1977: 117-118) posits *tl-. In this scenario, PT *k.t- simplified to /t-/ in Siamese and Yay but became *t.r- in Lungchow through weakening of *-t-. This intermediate cluster later simplified to *ʰr-, which is reflected now as /r-/ in Yay. In Long’an and a few other dialects, the modern reflex of *k.t- is /tʰ-/ which attests the intermediate step *tr-, i.e. *k.t- > *t.r- > *tr- > /tʰ-/ (see §4.4.4). 163 Table 4-23 Etyma with PT *k.t- Gloss ‘full’ ‘fart’ ‘big leaf’ ‘to wake up’ PT *k.temA *k.tɤtD *k.to:ŋA *k.tɯ:nB SI temA1 totDS1 tɔ:ŋA1 tɯ:nB1 LC timA1 tɤtDS1 to:ŋA1 tɯ:nB1 Y rimA1 ratDS1 roŋA1 rɯnB1 Others Long’an /tʰimA/ Long’an /tʰɯtDS1/ Long’an /tʰɤnB1/ In addition to *p.t- and *k.t-, other sesquisyllabic clusters with two stops can also be reconstructed. However, their reconstruction is rather tentative because they are attested only by one or two examples. These clusters include *p.q-, and *q.p-. Table 4-24 presents data for these clusters. These cases are viewed as aberrancies in Li (Li 1977). Table 4-24 Etyma with PT *p.q- and *q.p- Gloss ‘leg’ ‘to sell’ ‘body hair, feather’ PT *p.qa:A *p.qa:jD *q.pɯlA SI kʰa:A1 kʰa:jA1 kʰonA1 LC kʰa:A1 kʰa:jA1 kʰonA1 Y ka:A1 ka:jA1 punA1 Others Saek /kwa:A1/ Saek /kwa:jA1/ As discussed in §2.3.2, the labiality in Saek /kwa:A1/ and Saek /kwa:jA1/ is a retention from PT, as these two etyma still preserve the initial *p- in Sui /pa:1/ ‘leg’ and /pe1/ ‘to see’. Therefore, I posit *p.qa:A ‘leg’ and *p.qa:jA at the PT level respectively. The former has been related to PAN *paqa ‘thigh’ (Ostapirat 2005). As for ‘body hair’, it is not clear if the CT/SWT from *qon are related to the NT form 164 from *pɯl. If they are, they should be reconstructed as *q.pɯlA at the PT level. That is, SWT and CT simply dropped the non-initial *.p- but kept the labiality in the vowel. In contrast, NT lost the initial *q- when *q.p- simplified. Qinzhou, Yongnan, and Fusui show /pʰunA1/, /pʰɤnA1/, and /pʰɯnA1/ respectively. The aspiration in /pʰ-/ in these three dialects seems to be a trace of the initial *q.- 4.4.2 Voiceless obstruent + voiced stop PT had a number of sesquisyllabic clusters with voiceless obstruents followed by voiced stops. These C̥ .C̬ - clusters became simple implosives in most dialects, except for Saek where they became weakened to simple sonorants. Note that the Saek reflex of *C̥ .b- is /v-/ but this fricative must have relatively recently developed from *w-. The first set of data, presented in Table 4-25, can be reconstructed with *C̥ .b-. Table 4-25 Etyma with PT *C̥ .b- Gloss ‘shoulder’ ‘thin’ ‘light’ ‘side’ ‘to poison’ PT *C̥ .ba:B *C̥ .ba:ŋA *C̥ .bawA *C̥ .bɯəŋC *C̥ .bɯəA SI ba:B1 ba:ŋA1 bawA1 bɯəŋC1 bɯəA1 LC ba:B1 ba:ŋA1 bawA1 bɤ:ŋC1 bɤ:A1 Y ba:B1 ba:ŋA1 bawA1 Others Saek /va:B1/ Saek /va:ŋA1/ Saek /vawA1/ Saek /viəŋC1/ Saek /viəA1/ In these etyma, most dialects point unanimously to an earlier *ɓ- but Saek /v-/ suggests *ʰw-. This correspondence can be accounted for by positing PT *C̥ .b-. In this scenario, the medial *-b- in most dialects became an implosive *ɓ- when the voiceless 165 *C̥ - dropped out51, e.g. PT *C̥ .ba:B ‘shoulder’ > *ɓa:B > Siamese /ba:B1/ In contrast, the medial *.b- in Saek went through weakening and became *.w- before the voiceless *C̥ - was lost, e.g. PT *C̥ .ba:B ‘shoulder’ > * C̥ .wa:B > *ʰwa:B > Saek /va:B1/. Note that Ostapirat (2005) relates this etymon to PAN *qabaRa. In this account, the first consonant *q- in PAN *qabaRa corresponds to the initial *C̥ - in PT *C̥ .ba:B. A second set of data is given in Table 4-26. These etyma go back to PT *C̥ .d-. Table 4-26 Etyma with PT *C̥ .d- Gloss ‘black’ ‘bone’ ‘raw’ ‘catfish’ ‘mountain’ PT *C̥ .damA *C̥ .dukD *C̥ .dipD *C̥ .dokD *C̥ .do:jA SI damA1 du:kDL1 dipDS1 dukDS1 dɔ:jA1 LC damA1 du:kDL1 dipDS1 Y dokDL1 dipDS1 dokDS1 dojA1 Others Saek /ramA1/ Saek /rɔ:kDL1/ Saek /ripDS1/ Saek /rɔkDS1/ Saek /rɔ:jA1/ In parallel to *C̥ .b-, the PT sesquisyllabic cluster *C̥ .d- became an implosive *ɗ- in almost all dialects. Again, the only exception is Saek which has a weakened reflex *-r- in these etyma. For example, the non-initial *.d- in PT *C̥ .dipD became implosivized before the voiceless initial *C̥ - was lost, and is now reflected as /d-/ in Siamese, Lungchow, and Yay, e.g. *C̥ .dipD > *ɗipD > Siamese /dipD1/. In contrast, in Saek the non-initial *.d- went through weakening and became *.r- before the voiceless initial *C̥ - was lost, e.g. *C̥ .dipD > *C̥ .ripD > *ʰripD > Saek /ripD1/. Among the etyma 51 It is not clear what phonetic reasons underly the change from non-initial voiced stop preceded by a voiceless onstruent became implosive. However, this change is consistent with the fact that implosives are tonogenetically half-voiced, half-voiceless (see §6.2). 166 shown in Table 4-26, two have been claimed to have Austronesian connections (Ostapirat 2005): PT *C̥ .dipD ‘raw’ and PAN *qudip ‘alive’; and *C̥ .damA ‘black’ and PAN *tidem ‘black, dark’. These two Austronesian etyma lend support to the reconstruction of *C̥ .d- clusters for these etyma. In addition to *C̥ .b- and *C̥ .d-, a sesquisyllabic cluster with non-initial *.ɟmay have existed. The only potential case is ‘to stand’, which is reflected as /jɯ:nA1/ in Siamese and Lungchow, and as /dɯnA1/ in Yay. While the former points to earlier *ˀjɯ:nA, the latter suggests *ɗɯ:nA. These divergent developments probably reflect PT *C̥ .ɟɯ:nA. Specifically, the non-initial *.ɟ- became implosivized when preceded by a voiceless obstruent as expected but the resulting implosive was not *.ʄ-. Since *.ʄ- has never been allowed in Tai languages, in the SWT and CT dialects, implosive/glottalized palatal is precisely *ˀj-. As for NT, the non-initial *.ɟ- became implosive *ɗ- rather than *.ʄ- on account of this same ban against the palatal implosive. The development of implosives from plain voiced stops in the non-initial position strongly suggests that there was no contrast between medial voiced stops and their implosive counterparts. In other words, PT only distinguished plain voiced stops *b, and *d from implosives *ɓ and *ɗ in the initial position. 4.4.3 Voiced obstruent + voiceless stop As discussed in §3.4, a number of etyma with “voicing alternation” seem to go back to Proto-Kra-Dai. I have speculated that these items go back to sesquisyllabic clusters with voiced obstruents followed by voiceless stop in PT. These *C̬ .C̥ - clusters became aspirated in CT and SWT but became voiced in NT. However, it is not clear what the initial voiced consonants were. Most likely, the initial *C̬ - is different in different etyma. Etyma in Table 4-27 go back to PT *C̬ .t-. 167 Table 4-27 Etyma with PT *C̬ .t- Gloss PT ‘to arrive’ *C̬ .tɤŋA ‘to come into contact’ *C̬ .tɯ:kD ‘wild’ *C̬ .tɯəlB ‘to flood’ *C̬ .tuəmB SI tʰɯŋA1 thu:kDL1 tʰɯənB1 tʰuəmB1 LC Y tʰɤŋA1 taŋA2 thukDS1 tɯkDS2 tʰɤ:nA1 -t tɯənB2 tʰu:mB1 tumB2 Others The PT sesquisyllabic cluster *C̬ .t- was resolved differently in SWT and CT on one hand, and in NT on the other. In NT, the non-initial *.t- assimilated to *C̬ - and became voiced, e.g. PT *C̬ .tɤŋA > *dɤŋA > Yay /taŋA2/. In contrast, it became aspirated in SWT and CT dialects, i.e. *C̬ .tɤŋA > *tʰɤŋA > Siamese /tʰɯŋA1/. This etymon is probably related to PAN *dateŋ ‘to arrive’ (Thurgood 2007: 255) which shows the expected sequence *C̬ - + *.t-. Table 4-28 presents etyma that had *C̬ .k-, and *C̬ .q- at the PT level. Table 4-28 Etyma with PT *C̬ .k-, and *C̬ .q- respectively Gloss ‘son-in-law’ ‘pit’ ‘rice’ ‘to bite’ ‘excrement’ ‘bitter’ PT *C̬ .kɯəjA *C̬ .kumA *C̬ .qawC *C̬ .qɤpB *C̬ .qɯjC *C̬ .cɤkD SI kʰɤ:jA1 kʰumA1 kʰa:wC1 kʰopDS1 kʰi:C1 kʰomA1 LC kʰɤ:jA1 kʰumA1 kʰawC1 kʰopDS1 kʰi:C1 kʰumA1 Y kɯəjA2 kumA2 hawC2 hapDS1 hajC1 hamA2 Others 168 In parallel with *C̬ .t-, and *C̬ .c-, the PT sesquisyllabic cluster *C̬ .k- became a simple voiced *g- in NT but became voiceless aspirated *kʰ- in CT and SWT. For example, *C̬ .kɯəjA ‘son-in-law’ became *gɯəjA > /kɯəjA2/ in Yay but became *kʰɯəjA > /kʰɤ:jA1/ in Siamese. Similarly, PT *C̬ .q- became a simple voiced *ɢ- in NT but became voiceless aspirated *kʰ- in CT and SWT, merging with *C̬ .k-. For example, *C.qɯjC ‘excrement’ became *ɢɯjC > /hajC2/ in Yay but became *kʰi:C > /kʰi:C1/ in Siamese. 4.4.4 Voiceless stops + liquids/glides As discussed in §2.4.3, it is possible in a sesquisyllabic language to have a contrast between tautosyllabic clusters and sesquisyllabic clusters consisting of a stop followed by a liquid, e.g. *pr- vs. *p.r-. In PT, a few sesquisyllabic clusters with medial liquids or glides can be reconstructed. Among these clusters, clusters of type *C̥ .r- can be reconstructed with confidence. Table 4-29 presents example of PT *k.r-. This cluster corresponds to Li’s *kʰl- (Li 1977: 226-228). Table 4-29 Etyma with PT *k.r- Gloss ‘to imprison’ ‘to drive away’ ‘top for spinning’ ‘illness, fever’ ‘to beg’52 PT *k.raŋA *k.rapD *k.ra:ŋB *k.rajC *k.ro:A SI kʰaŋA1 kʰapDS1 kʰa:ŋB1 kʰajC1 kʰɔ:A1 LC haŋA1 ha:ŋB1 hajC1 ho:A1 Y caŋA1 ca:ŋB1 cajC1 Others Wuming /klapDS1/ Nandan /kjoA1/ 52 Gedney (n.d.) and Hudak (2008) have /kʰo:A1/ for Lungchow but Li (1940) and Zhang et al. (1999) has /ho:A1/. 169 For the etyma given above, Lungchow points to a voiceless liquid *ʰr- but Siamese and Yay suggest an earlier cluster *kr-, which did not come from PT *kr(see §4.2.1.4). This otherwise unexplained correspondence strongly indicates PT sesquisyllabic *k.r-, which either became a true cluster *kr-, or simply lost the initial *k- to become *ʰr-, depending on the language. For example, PT *k.raŋA became *kraŋA before aspirating to *kʰraŋA, which is now reflected as a simple /kʰaŋA1/ in Siamese. In contrast, the same etymon PT *k.raŋA became *ʰraŋA in Lungchow and is now reflected as /haŋA1/. Another sesquisyllabic cluster with a medial liquid that can be reconstructed with confidence is *p.r-, which contrasted with *pr- (see §4.2.1.1). In addition, Wuming /klapDS1/ ‘to drive away’ and Nandan /kjoA1/ ‘to beg’ shows that these two etyma are also found in NT. Li (1977: 86-87) reconstructs *pr- for this set of data. Table 4-30 presents etyma that go back to PT *p.r-. Table 4-30 Etyma with PT *p.r- Gloss ‘to be exposed’ ‘to burst’ ‘to hunt’ ‘cucumber’ PT *p.ra:kD *p.re:kD *p.rawB *p.rwe:ŋA SI LC Y ta:kDL1 pʰja:kDL1 ta:kDL1 tɛ:kDL1 pʰe:kDL1 tekDL1 pʰjawB1 tawB1 tɛ:ŋA1 pʰe:ŋA1 tiəŋA1 Others Saek /pra:kDL1/ Saek /pre:kDL1/ Saek /priəŋA1/ In the above data set, the Lungchow reflex /pʰj-/ points to earlier *pr-, but this *pr- must be secondary. This is because Saek shows unaspirated /pr-/ rather than the expected aspirated reflex /pʰr-/ for PT *pr- (see §4.2.1.1). Moreover, Siamese and Yay both show /t-/ indicating an earlier *t-. To account for this correspondence, I posit *p.r- which merged with PT *p.t- to give /t-/ in Siamese and Yay, and /pr-/ in Saek 170 (see §4.4.1). In parallel to *k.r- above, this PT *p.r- became reduced to a monosyllabic cluster and merged with PT *pr- in Lungchow, which now reflect this complex onset as /pʰj-/. Note that the etyma ‘cucumber’ have the speculative *-w- in addition to *p.rclusters, i.e. *p.rwe:ŋA (see §4.3). In addition to *k.r- and *p.r-, we may tentatively posit *t.r- and *c.r- for the following etyma: ‘to break’ and ‘to ask’ respectively. The etymon ‘to break’ is reflected as /hakDS1/, /tʰakDS1/, /tʰakDS1/, and /rak DS1/ in Siamese, Ningming, Long’an, and Yay respectively. These reflexes point to earlier *tr-, but the reflexes in many CT dialects point to *t-, e.g. Leiping and Lungchow /takDS1/. It is likely that PT *t.rmerged with *tr- earlier on in most dialects (see §4.2.1.2) but became simplified to *tin such CT dialects as Lungchow, etc. As for ‘to ask’, the reflexes in Siamese, Lungchow, and Yay are /tʰa:mA1/, /tʰa:mA1/, and /sa:mA1/, all pointing to *cr-. However, their counterparts in Debao, Jingxi, Western Nung, and Guangnan Nung are /kʰja:mA1/, /kʰja:mA1/, /cʰa:mA1/, and /cʰa:mA1/ respectively. I propose that in parallel to *t.r-, PT *c.r- merged with *cr- early on in most dialects, i.e. *c.r- > *cr- > /tʰ-/ or /s-/ (see §4.2.1.3). In Debao and like dialects, however, PT *c.r merged with *k.r- before both becoming *kr- > *kʰr- > /kʰj-/ (see §4.2.1.4). There is also evidence for sesquisyllabic clusters of the shape *C̥ .w-. This sesquisyllabic cluster corresponds to *ʰw- in Li’s system (1977: 82). Table 4-31 lists some examples of PT *q.w-. 171 Table 4-31 Etyma with PT *C̥ .w- Gloss ‘sweet, delicious’ ‘to sow’ ‘rattan’ PT *C̥ .wa:lA *C̥ .wa:lB *C̥ .wa:jA SI wa:nA1 wa:nB1 wa:jA1 LC va:nA1 va:nB1 va:jA1 Y va:nA1 va:jA1 Others Wuming /wa:nB1/ The etyma in this set of data all have tones in the first series and point to earlier *ʰw-. However, it is not possible to posit the voiceless glide for these cases because *ʰw- is regularly reflected as /f-/ in Siamese and Lungchow. Therefore, I posit a sesquisyllabic cluster *C̥ .w- to account for this correspondence. It is not clear what voiceless consonant the *C̥ .- might have been, but evidence from outside of Tai suggests *q.w- for ‘sweet’ and ‘rattan’. In particular, Proto-Lakkja (L-Thongkum 1992) has *kʰwa:nA for ‘sweet’. The initial *kʰ- seems to correspond to PT *q-, as Proto-Lakkja *k- corresponds regularly to Tai *k-. As for ‘rattan’, the reconstructed PAN form is *qúway, which also has *q- as initial. Therefore, it is possible that ‘sweet’ and ‘rattan’ had *q.w- as onset in PT. 4.4.5 Voiced consonant + liquid/glide In parallel to sesquisyllabic clusters of the type *C̥ .r- above, PT also had sesquisyllabic cluster with non-initial liquids or glides preceded by voiced consonants, including *C̬ .r-, and *C̬ .l-. Etyma that go back to these clusters all have tones in the second series, indicating that the initial consonants were voiced. Among these, *m.l- is the only cluster whose initial consonant can be recovered unequivocally. 172 Table 4-32 Etyma with PT *m.l- Gloss ‘slippery’ ‘grain’ ‘insect’ ‘body louse’ PT *m.lɯ:lB *m.lecD *m.le:ŋA *m.lelA SI lɯ:nB2 metDS2 mɛ:ŋA2 lenA2 LC lɤ:nB2 me:ŋA2 minA2 Y natDS2 neŋA2 nanA2 Others Saek /mlɯ:lB/ Saek /mlɛkDS2/ Saek /mlɛlA2/ The regular reflex of *m.l- in Siamese is /l-/ as in ‘slippery’ and ‘body louse’. The /m-/ in ‘seed’ and ‘insect’ is an instance of dialect borrowing as discussed in Pittayaporn (to appear-b) Except for *m.l-, there is no evidence for what their places of articulation or manners of *C̬ .- might have been. The first set of etyma, given in Table 4-33, go back to PT *C̬ .r-. Table 4-33 Etyma with PT *C̬ .r- Gloss ‘root’ ‘boat’ ‘day after’ ‘tomorrow’53 ‘bedbug’ PT SI *C̬ .ra:kD ra:kDL2 *C̬ .rwɯəA rɯəA2 *C̬ .rɯ:A rɯ:nA2 LC Y la:kDL2 ra:kDL2 lɯ:A2 ruəA2 lɯ:A2 rɯA2 Others *C̬ .rɯətD rɯətDL2 lɤ:tDL2 Longsheng /rɤ:tDL2/ 53 The Siamese form /-rɯ:nA2/ is now part of a monomorphemic word /ma.rɯ:nA2/ from an original compound /mɯ:C2 rɯ:nA2/. The /-n/ was acquired under the influence of the deictic /ni:C2/ ‘this’, because /ma.rɯ:nA2/ often occurs in the expression /marɯ:nA2 ni:C2/ ‘this coming day after tomorrow’. Also, see Li (1977: 128-129). 173 In the above data set, Siamese and Yay point to earlier *r- but Lungchow indicates *l-. To account for this discrepancy, Li (1977: 127-128) reconstructs *dr- for these etyma, but there is no evidence that the cluster had an initial *d-. In the current proposal, I propose that this set of etyma had sesquisyllabic clusters with a voiced initial followed by a non-initial liquid *.r-. The non-initial liquid *.r- in this environment merged with *.l- in the same environment. The result of this neutralization in Lungchow and many CT dialects is /l-/ while its reflex in Yay and other NT dialects is /r-/. The /r-/ in Siamese is a retention of the contrast between *C̬ .rand *C̬ .l-, as will be come clear in the discussion of *C̬ .l- below. Parallel to *C̬ .r-, PT also had *C̬ .l- clusters, exemplified by the etyma in Table 4-34. In these etyma, Siamese and Lungchow point to earlier *l- but Yay indicates *r. This is the same case of neutralization that we have just seen with respect to *C̬ .r-. In particular, *C̬ .r- completely merged with *C̬ .l- in CT and SWT but the reflexes are different in the two groups. Lungchow and many CT dialects show /l-/ as reflex of *C̬ .l- while Yay and other NT dialects has /r-/. Again, PT *C̬ .r- is also reflected as /l-/ in Lungchow and /r-/ in Yay. The contrast is preserved in Siamese and other SWT dialects where *C̬ .l- is reflected as /l-/, contrasting with /r-/ from PT *C̬ .r-. Table 4-34 Etyma with PT *C̬ .l- Gloss ‘to pull, drag’ ‘to steal’ ‘fingernail, toenail’ ‘wind’ ‘to lick’ PT *C̬ .la:kD *C̬ .lakD *C̬ .lepD *C̬ .lɯmA *C̬ .liəA SI la:kDL2 lakDS2 lepDS2 lomA2 liəA2 LC la:kDL2 lakDS2 lipDS2 lumA2 li:A2 Y ra:kDL2 rakDS2 rit DS2 -f rumA2 riəA2 174 Others Li (1977: 124-125) reconstructs *dl- for these etyma, but cognates in Lakkja indicate that the initial *C̬ .- was not the same in all etyma. The Lakkja forms for ‘to pull’, ‘to steal’, ‘fingernail’, and ‘wind’ are /kja:kD2/, /la:kD2/, /pli:pD1/, and /jɤmA2/ respectively. Although these Lakkja forms suggest different initial consonants for all of these etyma, the sound correspondence between Lakkja and Tai has not been worked out. Therefore, the sesquisyllabic clusters for this set of etyma are temporarily written as *C̬ .l-. 4.4.6 Clusters with non-initial nasals PT had a number of sesquisyllabic clusters with non-initial nasals. One of the most robustly supported is *t.n-. Almost all modern dialects show reflexes of *ɗ- for this set of etyma, but Saek unexpectedly has /tr-/. Li (1977: 129-131) included etyma in this set under his *ˀdr- together with etyma that are reflected as /ʰr-/ in Saek (see §4.4.2). Ferlus (1990) in contrast proposes that Saek /tr-/ came from original *ˀl- but believes that this glottalized liquid had become *ɗ- at the PT stage54. Although I agree with Ferlus that Saek /tr-/ represents a retention of a stage before the formation of *ɗ-, I claim that Saek /tr-/ goes back to a sesquisyllabic cluster *t.n- at the PT stage. Examples of etyma with PT *t.n- are given in Table 4-35. 54 Ferlus (1990: 12) does not reconstruct distinctions because of Saek’s marginal position within Tai. 175 Table 4-35 Etyma with PT *t.n- Gloss PT SI LC ‘star’ *t.na:wA da:wA1 da:wA1 ‘earthworm’ *t.nɯəlA dɯənA1 dɤ:nA1 ‘to transplant’ *t.namA damA1 damA1 ‘to be related by marriage’ *t.no:ŋA dɔ:ŋA1 ‘cockspur’ *t.nɤ:jA dɯəjA1 Y Others da:wA1 Saek /tra:wA1/ dɯənA1 Saek /truəlA1/ damA1 Saek /tramA1/ doŋA1 Saek /trɔ:ŋA1/ daA1 Saek /tra:A1/ The etyma in this set show that PT *t.n- coalesced and became *ɗ- in most dialects except for Saek, where the non-initial *.n- became *.r-. For example, PT *t.namA ‘to transplant’ became *ɗamA > /damA1/ in Siamese but *tramA > /tramA1/ in Saek. Therefore, Saek still preserves the original distinction between PT *ɗ-, *C̥ .d-, and *t.n- (see §2.3.2). PT *t.namA ‘to transplant’ seems to be related to the PAN form *taɬem ‘to plant’55, lending support for the reconstruction of PT *t.n- for this set of etyma. In addition to *t.n, PT also had clusters that consisted of other voiced stops followed by a non-initial *.n-. Examples of etyma with *C̬ .n- are given in Table 4-36. 55 This form corresponds to *taNem in Blust’s notation (Wurm & Wilson 1975). 176 Table 4-36 Etyma with PT *C̬ .n- Gloss ‘water’ ‘bird’ ‘bamboo shoot’ ‘dew, mist’ PT *C̬ .namC *C̬ .nokD *C̬ .na:ŋA *C̬ .na:jA SI LC Y na:mC2 namC2 ramC2 nokDS2 nukDS2 rokDS2 ra:ŋA2 na:jA2 Others Long’an /na:ŋA2/ Longsheng /rwa:jA2/ Although the tonal reflexes in modern forms unequivocally indicate that the the leftmost consonant of these *C̬ .n- clusters was voiced, it not clear exactly what consonants they were. Li (1977: 131-132) posit *nl/r- for these etyma but Lakkja data reveals clearly that the *C̬ .- in each etymon was different. Proto-Lakkja (L-Thongkum 1992) had *numC, *ʔmlokD, and *s-Na:ŋA for ‘water’, ‘bird’ and ‘bamboo shoot’, respectively. Furthermore, Ostapirat (2005) links the first two to PAN *daɬúm ‘water’56 and *manuk ‘bird’, which also show different initial consonants. Therefore, I temporarily write *C̬ .n- for these etyma. One exception is ‘bamboo shoot’, for which I posit *r.na:ŋA. The evidence for the initial *r- comes from some CT dialects such as Debao, Jingxi, Guangnan Nung, Western Nung. These varieties have either /r-/, /l-/, or /ð/, all of which point to *r- instead of the expected /n-/, cf. Debao /namC2/ ‘water’, /nɔkDS2/ ‘bird’ but /la:ŋA2/ ‘bamboo shoot’. 4.4.7 Other clusters In addition to the sesquisyllabic clusters discussed so far, some other combinations may have also existed in PT. Among these, some clusters consisted of two consonants, e.g. *r.t-, *t.h-, *q.s-, *m.p-, *s.c-, *z.ɟ-, *g.r-, *m.n- etc. Others may 56 This form corresponds to *daNum in Blust’s notation (Wurm & Wilson 1975). 177 have been more complex, e.g. *gm̩ .r-, *ɟm̩ .r-, *c.pl-, *g.lw- etc. However, they are sometimes attested by only one or two etyma. For example, the only case of *q.s- so far is *q.sepD ‘centipede’ which is reflected as Siamese /kʰepDS1/ (< PSWT *kʰrepDS1), Lungchow /kʰipDS1/, Jingxi /kʰjapDS1/. Qinzhou /lypDS1/, Wuming /θipDS1/, Du’an /θipDS1/. No other etymon shows this correspondence. Although one may argue that the etyma claimed to have these complex onsets in fact are not reconstructible, the fact that they are found throughout Tai and that they are usually regular regarding the rimes and the tones suggests that they might have been part of PT. Therefore, these clusters are tentatively reconstructed until more evidence becomes available. Refer to Appendix B for the reconstructions of individual etyma. Moreover, the inventory of reconstructed simple and complex onsets suggests that many more sesquisyllabic clusters may have existed. First of all, diverse types of sesquisyllabic onsets have been reconstructed for PT, but only a limited number of combinations have been reconstructed. For example, only one clusters consisting of a voiceless stop followed by a nasal has been reconstructed—*t.n-. Similarly, *k.r-, *p.r-, *t.r, and *c.r- have been reconstructed for sesquisyllabic clusters containing non-initial *-.r-, but *q.r- is still missing. Second, an abundance of voiceless nasals have been reconstructed for PT despite the presence of sesquisyllabic clusters. In SEA, voiceless sonorants are usually results of loss of a voiceless consonant at the left edge of tautosyllabic or sesquisyllabic clusters. For example, voiceless nasals in Burmese historically came from clusters with non-initial nasals, e.g. hna ‘face’ from *sna (Nishida 1975b). Therefore, it is very plausible that many etyma reconstructed with voiceless sonorants as onsets might in fact have had sesquisyllabic clusters. Taken together, these two facts about the inventory of reconstructed simple and complex onsets in PT seem to suggest that many more sesquisyllabic clusters existed but their traces have been lost in modern dialects. The Kam-Sui language 178 Lakkja seems to support this point of view. One crucial case is that of PT voiceless sonorants that correspond to Lakkja forms with initial /kʰ-/. These forms include pairs like PT *ʰma:A ~ Lakkja /kʰũəA/ ‘dog’, PT *ʰmu:A ~ Lakkja /kʰũ:A/ ‘pig’, PT *ʰmatDS1 ~ Lakkja /kʰũətD1/ ‘flea’, etc. This suggests that these etyma might have in fact been sesquisyllabic, i.e. *q.ma:A, *q.mu:A, and *q.matA respectively at the PT level. This would be perfectly in parallel with the reconstruction of *q.w- discussed in §4.4.4. However, in the current reconstruction, I opt to posit *ʰm- and *ʰr- for these etyma rather than *q.m- and *q.r- because Tai-internal evidence is lacking. 4.5 Reduction of sesquisyllabic clusters Even though I have argued that PT was a sesquisyllabic language, none of its daughter languages attested so far shows this typological profile. The biggest question is then what the processes through which PT sesquisyllables reduced to monosyllables were. The process of a sesquisyllabic language becoming monosyllabic can be referred to as “monosyllabicization.” In many languages of SEA, the monosyllabicization can be captured quite straightforwardly by a unified process in which sesquisyllables, disyllables, or polysyllables become reduced to monosyllables. Such “monosyllabification” is governed by a set of rules that determines the outcome of the reduction. For example, Tsat, an Austronesian language of Hainan Island, went through a monosyllabification process in which the first syllable was completely lost unless the medial was a liquid. 179 Table 4-37 Monosyllabification in Tsat (Sagart 1993) ‘eye’ ‘bee’ ‘fish’ ‘wet’ ‘white’ ‘plant’ ‘village’ ‘moon’ ‘new’ ‘shoulder’ Old Cham mata hani ʔikan pasah putih pala palei bulan birou bara Tsat ta2 ni2 ka:n2 sa1 ti1 pia2 (< pla) piei2 (< plei) pʰian3 (< blan) pʰiə (< brou) pʰia3 (< bra) The first set of data above shows that Old Cham disyllables were normally reduced to monosyllables by simply dropping the first syllables. In contrast, the second set of data shows that the resulting monosyllables still retained the initial consonants if they could combine with the medial consonants to form a cluster. A similar process of reduction must have operated in Tai languages but the details may have been different. Although one is also tempted to look for a set or sets of principles that make up a unified process of monosyllabification for Tai, comparative data in modern languages indicate that PT sesquisyllables had gone through various processes of cluster modification before the actual process of reduction took place. In this section, I outline a preliminary account of the reduction of PT sesquisyllables to monosyllables and illustrate how the reduction happened with case studies from four dialects: Siamese, Lungchow, Yay, and Saek. 180 4.5.1 Processes of cluster adjustment Unlike the case of Tsat, monosyllabicization in Tai did not involve only a process of monosyllabification. Rather, it is best characterized as a series of changes that ended with a process of monosyllabification. In other words, the sesquisyllabic PT had gone through sound changes that modified the original sesquisyllabic clusters before the monosyllabification of sesquisyllables finally took place. The sound changes that resulted in adjustments of sesquisyllabic clusters prior to the monosyllabification include weakening, implosivization, metathesis, assimilation, and simplification. The process of weakening can be characterized as a change in which a segment becomes less consonant-like (Trask 2000: 190). Many of the PT sesquisyllabic clusters underwent weakening. For example, the non-initial *.t- in PT *p.t- was weakened to *.r- in many dialects including Lungchow and Saek (see §4.4.1). The result of this weakening of *.t- is that PT *p.t- became *p.r-. In Lungchow, the resulting *p.r- clusters lost the initial *p- and became *ʰr-. This *ʰr- merged with the original *ʰr- and is now reflected as /h-/ with tones from the first series. In Saek, the *p.r- cluster that resulted from the weakening merged with the original PT *p.r- so that both PT *p.r- and *p.t- are now reflected as a monosyllabic cluster /pr-/. Another example of weakening is that of PT non-initial voiced stops in *C̥ .C̬ - clusters (see §4.4.2). In Saek, a non-initial *.d- that occurred after an initial voiceless stop was weakened to *.r-. The resulting cluster *C̥ .r- later went through monosyllabification and became *ʰr-, which is now reflected as /r-/. These cases of weakening are schematized in Figure 4-1. 181 PT *p.ta:A ‘eye’ *p.ra:A *t.ra:A *pra:A Lungchow /ha:A1/ Saek /pra:A1/ PT *C̥ .dipD ‘raw’ *C̥ .rip *ʰripD Saek /ripDS1/ Figure 4-1 Weakening of *p.t- and * C̥ .d- The second process of cluster adjustment is implosivization, a change in which a segment becomes implosive. In Tai varieties, implosivization occurred mainly in two types of clusters: *t.n- and *C̥ .d-. In the former, the two consonants in PT *t.n- were collapsed together yielding an implosive *ɗ- in almost all dialects. The only language that did not undergo the implosivization is Saek, in which the non-initial *.nweakened to /-r-/ (see §2.3.2). In *C̥ .d-, PT non-initial *.d- became implosivized to PT *.ɗ-. Later the resulting cluster *C̥ .ɗ- was simplified to simple *ɗ-. Again, all Tai dialects except for Saek went through this change. The non-initial *.d- was weakened to *.r- in Saek (see §4.4.2). The implosivization of *t.n- and *C̥ .d- is schematized in Figure 4-2. 182 PT *t.namA ‘to transplant’ *ɗamA *t.ramA Siamese /damA1/ Saek /tramA1/ PT *C̥ .dipD ‘raw’ *C̥ .ɗipD *ɗipD *C̥ .ripD Siamese /dipDS1/ Saek /ripDS1/ Figure 4-2 Implosivization of *t.n- and *C̥ .d- The third process of cluster adjustment is metathesis, which refers to a change in which the order of segments in a word is altered (Trask 2000: 211). Two examples of PT sesquisyllabic clusters that were metathesized are *r.t- and *p.q-. The PT cluster *r.t- can be found in the etymon *r.tɯəB ‘sweat’. It was metathesized to *t.r- in SWT and most CT dialects. The resulting cluster merged with PT *t.r- which later became monosyllabified to *tr-. Now the cluster is reflected as /h-/ in both Siamese and Lungchow (see §4.2.1.2 and §4.4.4). Shangsi and a few other CT dialects did not metathesize but voiced the non-initial *.t- under the influence of *r.-, yielding *r.d-. This intermediate cluster was later reduced and became *d-, e.g. Shangsi /tyB2/ from *r.dɯəB < PT *r.tɯəB. The second example is the development of PT *p.q- in Saek. In this NT language, PT *p.q- is reflected /kw-/, suggesting that *p.q- metathesized to *q.p- before the non-initial *.p- was weakened to *.w-. In other dialects, the initial *p.- simply dropped out due to the monosyllabification. The metathesis of *r.t- and *p.q- is schematized in Figure 4-3. 183 PT *r.tɯəB ‘sweat’ * t.rɯəB PT *p.qa:A ‘leg’ *q.pa:A *q.wa:A Figure 4-3 Metathesis of *r.t-, and *p.q- Siamese /hɯəB1/ Saek /kwa:A1/ The fourth process to be discussed is assimilation, which refers to a change in which a segment becomes more similar to another segment in the same phonological sequence (Trask 2000: 30). An example of assimilation in PT sesquisyllabic clusters is the case of PT *p.t-. In Debao and a number of other CT dialects, this PT cluster went through weakening so that its non-initial *.t- became *.r-, resulting in a cluster *p.r-. This new *p.r- clusters subsequently went through assimilation such that its *p.assimilated to *.r- resulting in *t.r-. This *t.r- merged with the original PT *tr- and is now reflected as /tʰ-/ in Debao (see §4.2.1.2 and §4.4.4). PT *p.ta:A ‘eye’ *p.ra:A *t.ra:A Figure 4-4 Assimilation of *p.t- *tra: A Debao /tʰa:A1/ The last process that targeted PT sesquisyllabic clusters is simplification, which refers to the reduction of a consonant cluster by dropping one or more of its constituents (Trask 2000: 310). An example of simplification in Tai occurred with PT 184 *t.h- in *t.hajC ‘to weep’. In all NT and a few CT dialects, the cluster *t.h- lost its noninitial *.h- and became a simple *t-, merging with the original PT *t-. In Yay, *t.hajC is now reflected as /tajC1/. Note that the process of monosyllabification is essentially a simplification process although for purposes of the discussion it is treated as a distinct process in this account. The simplification of *t.h- is schematized in Figure 4-5. PT *t.hajC ‘to weep’ *tajC Yay /tajC1/ Figure 4-5 Simplification of *t.h- In this subsection, I have shown that at least five processes of cluster adjustment are commonly found among Tai languages. It is not the case that all dialects underwent all of these processes. Nor is it true that all dialects went through the same subsets of processes. In the following subsection, I discuss the process of monosyllabification which reduced the sesquisyllabic clusters that had undergone the cluster modification to monosyllabic clusters. 4.5.2 Monosyllabification of sesquisyllables After undergoing the processes of cluster adjustments described above, PT sesquisyllables subsequently went through a process of monosyllabification. The general result of this process was that sesquisyllabic clusters were reduced to true tautosyllabic clusters. Like the case of Tsat briefly discussed earlier, the monosyllabification of Tai languages can be captured with a set of developments that determines the outcome of the cluster simplification. These developments conspire so 185 that the results of the reduction obeyed the Sonority Sequencing Principle (see §2.2.2.2). The first development involves sesquisyllabic clusters with medial liquids. A PT sesquisyllabic *C̥ .l- or *C̥ .r- cluster whose initial *C̥ - was a voiceless stop kept both consonants but became a tautosyllabic cluster. For example, PT *k.r- became *kr- in SWT and NT when the monosyllabification took place. PT *k.r- became *krand then aspirated to PSWT *kʰr-. It is now reflected as /kʰ-/ in Siamese. In some CT dialects, PT *k.r- had undergone simplification and became *ʰr- before the monosyllabification took place. In Lungchow, it is now reflected as /h-/ (see §4.4.4). Another example is PT *p.r- which became *pr- in Lungchow when the monosyllabification took place, merging with the original *pr-. Subsequently, aspiration of clusters with medial *-r- took place and transformed it to *pʰr-, which is now reflected as /pʰj-/ (see §4.4.4). The monosyllabification of *k.r- and *p.r- is schematized in Figure 4-6. PT *k.raŋA ‘to imprison’ *ʰraŋA *kraŋA Lungchow /haŋA1/ Siamese /kʰaŋA1/ PT *p.ra:kD ‘to be exposed’ *pra:kD Lungchow /pʰja:kDL1/ Figure 4-6 Monosyllabification of *k.r- and *p.r- The second development determines more generally which consonant within a cluster was dropped. As a rule, the first syllable of a *C.C- cluster is dropped while the 186 second is retained. For example, PT *k.t- went through two lines of development. NT dialects went through the first one which involved weakening of non-initial *.t-, i.e *k.t- > *k.r-. CT and SWT, on the other hand, did not go through weakening and remained *k.t-. Therefore, at the time of monosyllabification, PT *k.t- had two distinct reflexes: *k.r- in NT and *k.t- in CT and SWT. According to this first principle, the second consonant of each of their reflexes was kept while the first was dropped. Therefore, NT *k.r- was monosyllabified to *ʰr-, but CT and SWT *k.t- went to *t-. The *ʰr- is now reflected as /r-/ in Yay (see §3.6.9.1) while the *t- is reflected as /t-/ in Siamese and Lungchow (see §3.6.1.2). Another illustration of this development is the case of *t.h-. This PT sesquisyllabic cluster went through simplification and became *t- in Long’an but remained *t.h- in CT and SWT. At the time of monosyllabification, the *t.h- preserved in SWT and CT was simplified to *h- as this fricative is the second consonant of the cluster. In contrast, the *t- in Long’an did not go through monosyllabification because it was already monosyllabic. The development of these two clusters can be schematized in Figure 4-7. PT *k.temA ‘full’ PT *t.hajC ‘to cry’ *k.remA *tajC *ʰremA *temA Yay /rimA1/ Siamese /temA1/ Long’an /tajC1/ Siamese /hajC1/ Figure 4-7 Monosyllabification of *k.t- and *t.h- 187 The last development determines the phonation type of the resulting clusters. Like many other languages of SEA, it is the phonation type of the first consonant that is preserved on the resulting clusters, cf. Chamic (Brunelle 2008; Thurgood 1999). For example, PT *p.r- became *p.t- in SWT and some NT dialects. The resulting *p.tcluster was eventually monosyllabified to a voiceless *t- because both *p- and *twere voiceless. In contrast, PT *C̥ .d- became *C̥ .r- in Saek through weakening (see §4.4.2). At the time of monosyllabification, the initial *C̥ - was lost but its phonation type was preserved on the resulting *ʰr. Therefore, modern Saek has /r-/ with the first series of tones for PT *C̥ .d-, reflecting the voicelessness of the initial *C̥ . Another example is the case of the voicing-alternation items (see §3.4 and §4.4.3). Voicing alternation items that had *d- in an earlier stage of NT came from PT *C̬ .t-. At the time of monosyllabification, the voiced *C̬ - was dropped but the non-initial *.t- took on its voicing and became *d-. The reflex of PT *C̬ .t- in Yay is /t-/ with tones in the second series, reflecting the phonation type of the first consonant in PT *C̬ .t-. The developments of *p.t-, *C̥ .d-, and *C̬ .t- are schematized in Figure 4-8. PT *p.ta:A ‘eye’ *ta:A Siamese /ta:A1/ PT *C̥ .dipD ‘raw’ *C̥ .ripD *ʰripD Saek /ripDS1/ PT *C̬ .tɤŋA ‘to arrive’ *dɤŋA Yay /tɤŋA2/ Figure 4-8 Monosyllabification of *p.t-, *C̥ .d-, and *C̬ .t- 188 In this subsection, I have discussed the monosyllabification of PT sesquisyllables. I have shown that it is not the original PT sesquisyllabic clusters that underwent the monosyllabification because the process occurred after a number of sound changes, e.g. weakening, assimiliation, simplification etc., had taken place (see §4.5.1). Therefore, the sesquisyllabic clusters that were materials for the monosyllabification were the result of those changes. The monosyllabification itself consisted of three developments that work together to transform sesquisyllables to monosyllables whose onset obeys the Sonority Sequencing Principle. 4.6 Summary In this chapter, I have discussed the two types of PT complex onsets: tautosyllabic clusters and sesquisyllabic clusters. I have shown that tautosyllabic clusters in PT can consist of up to three consonants and must obey the Sonority Sequencing Principle. Sesquisyllabic clusters, on the other hand, usually violated the Sonority Sequencing Principle. In modern Tai dialects, PT sesquisyllables have been reduced to monosyllables through a long process of reduction. I have shown that this process cannot be viewed as a unified process of monosyllabification because PT sesquisyllabic clusters had undergone various phonological changes before they were finally reduced to tautosyllabic clusters. In sum, this chapter is the first systematic attempt at positing sesquisyllables for PT and their developments into monosyllabic words in modern Tai languages. 189 CHAPTER 5 PROTO-TAI RIMES 5.1 Introduction Perhaps the least understood aspect of PT reconstruction is its vocalic system. The central problem seems to be that the number of correspondence sets that have been identified is so great that one is led to account for it by positing a complex vocalism (Li 1977: 297-299). While Li (1977) holds that PT vowel system lacked vowel length but consisted of a large number of diphthongs and triphthongs, Sarawit (1973) posits a vowel-length contrast and thus a smaller set of complex vowels. This problem has also lead Haudricourt (1975) and Strecker (1983) to envision PT as a disyllabic language with a relatively simple vowel system. In contrast to the vowels, the system of final consonants is by consensus judged to be straightforward. This view is reflected in the work of Li (1977: 58) who states that the PT final system is preserved with very little change in most dialects. The inventory of codas that Li proposes includes only stops, nasals, and glides, with a possible liquid *-l as attested in Saek. However, approaching these two components of PT phonology separately is probably ill-conceived. Given that vowels and codas frequently function together as one unit cross-linguistically, it is highly likely that their mutual influence is at work in the observed correspondence patterns among modern languages. In this study, the PT vowels and codas are treated together as rimes. In Chapter 2, I established that PT was a sesquisyllabic language in which prosodic words can be either monosyllabic or sesquisyllabic. Due the maximality and minimality constraints, each PT prosodic word must be exactly one iambic foot. In this chapter, I propose an inventory of PT rimes made up of a comparatively small number of vowels and a relatively rich set of final consonants. Three aspects of the rime 190 system are discussed: 1) the reconstruction of the vowel-length contrast, 2) the solution to the so-called “Gedney’s Puzzle,” and 3) the existence of final*-l, *-c, and possibly *-ɲ. Lastly, I present the reconstructions of each PT rime. 5.2 Inventory of PT rimes In §2.3, I argued that PT is best characterized as a sesquisyllabic language, which allowed both simple monosyllabic and sesquisyllabic words. Obviously, each monosyllable can only have one rime. This is also the case for sesquisyllables because the first “half-syllable” allows neither a contrastive vowel, nor a real coda consonant, as discussed in §2.4.3 and §2.5. Therefore, the rime of a sesquisyllable refers more specifically to the rime of the major syllable. As outlined in §2.5.1.2, size requirements on the prosodic word dictate that each PT rime must contain exactly two timing units, allowing for long open rimes (-V:), short closed rimes (-VC), and long closed rimes (-V:C). Examples of words allowed by the PT phonotactics are *ka:A ‘crow’, *katD ‘to take a bite’, and *ka:tD ‘mustard green’, but not *kaA. Because PT rimes cannot take up more than two timing units, the coda in CV:C syllables likely share the second mora with the vowel as is the case in Siamese as analyzed by Morén and Zsiga (2006). I propose that the PT vowel system consists of 7 simple vowels, each occurring as long or short, and five diphthongs. The PT inventory of simple vowels is given in Table 5-1. This triangular system is typologically well-attested (Maddieson 1984b). 191 Table 5-1 Inventory of PT simple vowels front back unrounded unrounded rounded high *i, *i: *ɯ, *ɯ: *u, *u: mid *e, *e: *ɤ, ɤ: *o, *o: low *a, *a: Overall, this inventory is identical to that proposed by Sarawit (1973) but differs substantially from the conventional reconstruction. Two aspects that should be highlighted are the contrastive vowel length and the three-way height contrast. For each vowel quality, both short and long vowels were contrastive. As far as vowel height is concerned, a vowel can be either high, mid, or low. The low series consists only of a long and short *a: and *a. Turning to complex vowels, the inventory of PT diphthongs is given in Table 5-2. Table 5-2 Inventory of PT diphthongs rising falling front back unrounded unrounded rounded *iə *ɯə *uə *ɤɰ *aɰ An outstanding characteristic of the proposed diphthong inventory illustrated above is its relative small size, especially when compared to the system of Li (1977) and Sarawit (1973). This system consists of three rising diphthongs and two falling diphthongs. These diphthongs do not contrast for length. The lack of length contrast in 192 diphthongs is expected because long diphthongs would exceed the two timing units allowed. As short diphthongs have two timing units by definition, long diphthongs would have to be linked to three timing units. Except for *ɤɰ and *aɰ posited here, surface rising diphthongs are treated as vowels plus glides phonologically because they do show length contrast. For example, *[ai] is phonologically a vowel *a followed by *-j contrasting with *[a:i], which is phonologically *a:j. Of the PT consonants that can occur in onset position, fewer than half can function as codas. In each syllable, only one final consonant is allowed. The inventory of PT codas is given in Table 5-3. Table 5-3 Inventory of PT codas stop fricative nasal liquid glide labial *p *m *w alveolar palatal *t *c *n (*ɲ) *l *j velar *k *ŋ uvular glottal The set of PT final consonants can be characterized as the onset inventory reduced by synchronic positional neutralizations. In terms of manners of articulation, PT did not distinguish between fricatives and stops so that the only final obstruents allowed were voiceless stops. As for sonorants, only nasals and the lateral *-l could occupy the coda slot. This manner neutralization can be understood as a neutralization of the feature [+/-continuant] found widely across languages. In terms of places of articulation, the dorsal places of articulation were extremely impoverished when 193 compared to onset position. Although velar, uvular, and glottal consonants could all function as onsets, only the velar stop and nasal can be in coda position. This can be considered another case of positional neutralization. As for phonation type, most noticeable is the absolute lack of contrast. While PT onsets can be either voiceless, implosive/glottalized, or voiced, only one phonation-type is possible for each manner of articulation in coda position: all final obstruents are voiceless but all final sonorants are voiced. This can be understood as laryngeal neutralization. This generalization agrees with Li’s and Sarawit’s reconstructions. This restricted inventory of coda consonants is typologically very common and is one of the most widespread SEA areal traits (Matisoff 2001; Rhee 2003). Cross-linguistically, neutralizations of manner, place, and laryngeal features are robustly attested. 5.3 PT vowels: quality and quantity The question whether there was a quantity contrast among PT vowels is one of the most crucial factors in understanding PT phonology. This question also ties very closely to the issue of vowel qualities, especially vowel heights. Opinions among researchers differ with regard to whether both quality and quantity were contrastive in PT vocalism. While Li (1977) and Luo (1997) consider vowel length to have developed secondarily in daughter languages, Haudricourt (1948; 1966-1974), Wang (1966), Nishida (1954; 1955), and Sarawit (1973) hold that it was contrastive in PT. In this section, I argue in favor of the latter position. As a point of departure, the inventories of PT simple vowels according to Li and Sarawit are illustrated in Table 5-4. 194 Table 5-4 PT vowel system according to Li (1977) and Sarawit (1973) *i *ɯ *u *e *ɤ *o *ɛ *a *ɔ *i, *i: *e, *e: *ɯ, *ɯ: *ɤ, ɤ: *a, *a: *u, *u: *o, *o: Li (1977) Sarawit (1973) Li (1977) proposes a PT vowel system that distinguished three degrees of vowel height but lacked vowel length contrast. He argues that in many dialects long and short vowels do not contrast, except for *a and *a:. He interprets the contrast in the low vowel as one of quality rather than one of quantity (Li 1977: 259). Taking Siamese as a case study, he further argues that the quantity contrast among non-high vowels was introduced by borrowing or secondary lengthening. He only recognizes Siamese short mid vowels /e/, and /o/, together with long low vowels /ɛ:/ and /ɔ:/ as having PT origins. Although he admits that the length contrast among high vowels in Siamese does point to a distinction in PT, he argues that Siamese long high vowels /i:/, /ɯ:/, /u:/ resulted from monophthongization of PT *ie, *ɯe, *ue. Luo (1997: 56-63) assumes Li’s position that PT did not have a vowel-length contrast but also argues that the low vowels *ɛ and *ɔ did not exist. Although he correctly identifies some cases of low vowels in modern dialects as secondary developments, he fails to recognize that many of these cases may have been conditioned by vowel length. For example, he claims that PT *e and *o became *je and *wo in closed syllables in the SWT dialect of Tai Ya, spoken in the upper reaches of the Red River, e.g. /kjemC1/, and /swoŋA1/ from what he would reconstruct as *kemC ‘cheek’57 and *soŋA ‘two’58 Although Luo is correct in pointing out that the 57 Reconstructed as *kɛmC in Li’s system and as *ke:mC in the current reconstruction. 195 breaking seen in Tai Ya is secondary, he ignores the fact that there are equally many cases of closed-syllable items that show simple /i/ and /o/ (presumably from earlier *e and *o respectively), e.g. /mitDS2/ and /hokDS1/ from what he would also have to set up *mletD ‘grain’59 and *xrokD ‘six’60. Moreover, he does not discuss cases of minimal pairs such as /pitD1/ ‘duck’ from PT *petD vs. /pjetD1/ ‘eight’ from *pe:tD, and /nokD1/‘bird’ from PT *C̬ .nokD vs. /nwokD1/ ‘outside’ from PT *l̩ .no:kD. It is impossible to explain what conditioned the alleged split if the two sets of words come from the same PT sources61. Prominent proponents of contrastive vowel length in PT are Haudricourt (1948; 1966-1974), Sarawit (1973), Nishida (1954; 1955) and Wang (1966), among others. Although Haudricourt does reconstruct contrastive short and long vowels, he did not provide a picture of the entire system nor discuss it explicitly. The next major study to posit a primary length contrast in the PT vowel system is Sarawit (1973) who meticulously reconstructs PT vocalism bottom-up by first reconstructing the intermediate proto-languages of SWT, CT, and NT. In contrast to Li, Sarawit proposes that the low vowels /ɛ:/ and /ɔ:/ in Siamese and other dialects resulted from lowering of long mid vowels */e:/ and */o:/ respectively62. Furthermore, she views the quantity contrast among high vowels as a retention from PT rather than the result of simplification of original diphthongs as Li proposes. Although Li’s and Sarawit’s views are both appealing, only the latter can account for the distributional gaps in PSWT. Elsewhere (2008b; to appear-b) I have shown that both length and height were contrastive in PSWT only for high vowels: the 58 Reconstructed as *klɔŋA in Li’s system and as *klo:ŋA in the current reconstruction. 59 Reconstructed as *mletD in Li’s system and as *mlecD in the current reconstruction. 60 Reconstructed as *xrokD in Li’s system and as *krokD in the current reconstruction. 61 I believe that the change from PT to Tai Ya is as follows: PT */e:/, */o:/ > PSWT */ɛ/, */ɔ/ > /je/, /wo/. 62 In contrast to Luo, she correctly posits this development only for long vowels. 196 distribution of height in mid and low vowels is predictable. Specifically, the PSWT mid vowels *e, *ɤ, and *o are always short and cannot occur in open syllables, but the low vowels *ɛ: and *ɔ: are always long and can occur in open syllables. This distribution is illustrated in Table 5-5. Table 5-5 Distribution of PSWT non-high vowels open syllables long Short mid low *ɔ: closed syllables long short *ok *ɔ:k The distributional pattern in Table 5-5 argues against Li’s account and lends strong support to Sarawit’s proposal for PT. That is, Li cannot account for why the PSWT mid vowels *e, and *o did not occur in open syllables. In his scenario (Li 1977), PT *ɛ and *ɔ were long because they were subject to regular open-syllable lengthening. In the same way, PT *e and *o were not lengthened because they did not occur in open syllables. However, it is not clear why PT mid vowels *e and *o were banned from open syllables in the first place. Sarawit (1973)’s system, in contrast, offers a very powerful explanation for these distributional gaps. Although she also believes that PT vowels in open syllables are always long, she holds that there was a quantity contrast in PT closed syllables. It is precisely this quantity contrast that explains why short *e and *o were not allowed in open syllables. That is, only long vowels were allowed in open syllables. This ban against short vowels in open syllables is exactly an effect of word minimality discussed in §2.2.2.1. Word minimality dictates that a prosodic word must be at least 197 one iambic foot. An open syllable with a short vowel is thus not allowed because it is light and cannot head an iambic foot. In this scenario, PT long mid *e: and *o: lowered and became PSWT low vowels *ɛ: and *ɔ:, retaining their original length. Because PT vowels in open syllables were all long, all mid vowels in open syllable became low vowels in PSWT. In contrast, their short counterparts *e, and *o did not lower and were kept as short mid vowels. Because PT short vowel only occurred in closed syllables, there was no PT source for PSWT mid vowels in open syllables. The lack of PT *e and *o thus results in the observed gaps in PSWT. Therefore, these otherwise peculiar gaps are strong evidence for quantity contrast in PT. Dhananjayananda (1997)’s study illustrates how these gaps were filled in Siamese. Not only does this system of contrast nicely account for the comparative data, but it is also consistent with the typology of the prosodic word in SEA. As shown in §2.4.3, sesquisyllabic languages impose a bimoraic minimum on the prosodic word so that all monosyllables and the major syllables of sesquisyllabic words are heavy. This is exactly the system reconstructed for PT. Furthermore, the proposed vowel inventory is a very likely one from a typological point of view. More specifically, the vowel system posited here is a triangular system, which has been shown to be typologically more common than square systems like that proposed by Li (Maddieson 1984b). 5.4 Solving Gedney’s Puzzle First discussed in detail by Gedney (1972b), the so-called “Gedney’s Puzzle” refers to particular sets of vowel correspondences involving high vowels. A number of Tai etyma unequivocally go back to PT simple vowels *i, *ɯ, and *u in Li’s system, which correspond to long *i:, *ɯ:, and *u: in the system proposed here. Another set of etyma clearly go back to PT rimes *aj, *aw, and *aɰ. In addition to these two sets, 198 there are a number of items that seem to point to earlier *i:, *u:, and *ɯ: in some languages but *aj, *aw, and *aɰ in others. The reverse patterns also exist. To illustrate the issue, first examine the two sets of data provided in Table 5-6 and Table 5-7. Table 5-6 Items that unequivocally go back to PT */i:/, */ɯ:/, and */u:/, i.e. the ‘year’ set Gloss ‘year’ ‘four’ ‘day’ ‘book’ ‘pig’ ‘door’ Li Siamese pi:A1 *i si:B1 mɯ:C2 *ɯ sɯ:A1 mu:A1 *u tu:A1 Lungchow pi:A1 ɬi:B1 mɯ:C2 ɬɯ:A1 mu:A1 tu:A1 Yay piA1 θiB1 fɯC2 θɯA1 muA1 tuA1 In Table 5-6 above, Siamese, Lungchow, and Yay represent SWT, CT, and NT respectively. These three varieties all have /i:/, /u:/, and /ɯ:/ for these correspondences. However, some varieties, e.g. Debao and Lungming, have diphthongized their original long high vowels, e.g. Lungming /pejA1/ ‘year’, /sɤɰA1/ ‘book’, and /mowA1/ ‘pig’. Having established that vowel quantity was contrastive in PT, it is clear that the data provided above points to original simple long high vowels. Li reconstructs *i (Li 1977: 262-263), *ɯ (Li 1977: 264-265), and *u (Li 1977: 266267) for this set of etyma. For convenience, this set of data may be referred to as the ‘year’ set. 199 Table 5-7 Items that unequivocally go back to PT */aj/, */aɰ/, and */aw/, i.e. the ‘egg’ set Gloss ‘egg’ ‘to cough’ ‘leaf’ ‘clear’ ‘old’ ‘head louse’ Li Siamese kʰajB1 *ɤj ɁajA1 bajA1 *ɤɰ sajA1 kawB1 *ɤw hawA1 Lungchow kʰjajB1 ɁajA1 baɰA1 ɬaɰA1 kawB1 hawA1 Yay rajB1 ɁajA1 baɰA1 θaɰA1 kawB1 rawA1 In Table 5-7, all the three varieties unequivocally show reflexes of earlier *aj, *aɰ, and *aw for the etyma above. In dialects where PT high vowels diphthongized, the contrast between the two sets of PT rimes is still preserved. For example, Debao has /ɤj/ for PT *i: but /aj/ for PT *aj, cf. /pɤjA1/ from PT *pi:A ‘year’ and /pajA1/ from *pajA ‘to go’. Note that in Li’s system the rimes of ‘to cough’ and ‘egg’ are reconstructed as *ɤj (Li 1977: 285-286), the rime of ‘leaf’ and clear’ as *ɤɰ (1977: 288-289), and the rime of ‘old’ and ‘head louse’ as *ɤw (Li 1977: 290). This set of data may be referred to as the ‘egg’ set. In contrast to the two sets shown so far, for other items, modern dialects seem to give conflicting evidence on how the PT rimes for these words should be reconstructed. 200 Table 5-8 The ‘fire’ set of Gedney’s Puzzle correspondences Gloss ‘fire’ ‘dry field’ ‘breath’ ‘dry’ ‘empty’ ‘nine’ Li *ɛj *eɰ/ɛɰ Siamese fajA2 rajB2 cajA1 *jɤw pla:wB1 ka:wC1 Lungchow fajA2 ɬajB2 caɰA1 kʰaɰB1 pjawB1 kawC1 Yay fiA2 riB2 sɯA1 pjuB1 kuC1 We may refer to the data in Table 5-8 as the ‘fire’ set. Siamese lacks ‘dry’ but White Tai has /xaɰB1/. Similarly, Yay does not have this form but Po-ai has /hɯ:B1/. It is generally held that the data in this set pattern with the ‘egg’ set in SWT and CT, but pattern with the ‘year’ set in NT. That is, the CT and SWT reflexes for the etyma in this set seem to suggest complex rimes *aj, *aɰ, and *aw in PT but their NT counterparts seem to go back to simple rimes *i:, *ɯ: and *u:. Li reconstructs *ɛj (Li 1977: 286) for ‘fire’ and ‘dry field’, *ɛɰ or *eɰ for ‘breath’ and ‘dry’ (Li 1977: 289), and *jɤw for ‘empty’ and ‘nine’ (Li 1977: 291). To complicate the situation, the reverse pattern also exists. 201 Table 5-9 The ‘long’ set of Gedney’s Puzzle correspondences Gloss Li Siamese ‘long’ ‘excrement’63 ri:A264 *ei kʰi:C1 ‘crab’ pu:A1 ‘paternal grandfather’ *ew pu:B1 Lungchow ɬi:A2 kʰɤjC1 pu:A1 Yay rajA2 hajC2 pawA1 The set of etyma in Table 5-9 may be referred to as the ‘long’ set. CT lacks the etymon ‘paternal grandfather’ altogether while some NT languages still retain it, cf. Po-ai /pawB1/. The received generalization is that it patterns with the ‘year’ set in SWT and CT, but with the ‘egg’ set in NT. In other words, the CT and SWT reflexes of etyma belonging to this set point to earlier *i:, *ɯ:, and *u: but their NT reflexes suggest *aj, *aɰ, and *aw. Li posits *ej for ‘long’, and ‘excrement’ (Li 1977: 286), and *ew for ‘crab’ and ‘paternal grandfather’ (Li 1977: 291). Notice that the pattern /ɯ/~/aɰ/, though expected, is not attested. As Gedney argues (1972b), it is clear that this puzzle does not indicate doublets in PT, since no language is known to have inherited both the monophthongal and diphthongal forms for any given etymon. However, it remains unclear what these data indicate about the rime system of PT. Confronted with such puzzling sets of data, various researchers have proposed different accounts. Li (1977), followed by Liang and Zhang (1996) accounts for these “Gedney’s Puzzle” items by positing very fine contrasts among complex vowels as we have seen in our discussion of the data sets. Sarawit (1973) similarly posits complex vowels for these etyma but differs in the actual forms posited. In contrast, Haudricourt (1975), followed by Strecker (1983), 63 This is a ‘voicing alternation’ item. See §3.4. 64 This form means ‘oval’ in Thai 202 proposes that this is a kind of “umlaut” phenomenon. Specifically, the vowel of the first syllable in a disyllabic etymon affected the vowel of the second syllable differently in the process of monosyllabicization. Li’s, Sarawit’s, and Haudricourt’s solutions are compared in Table 5-10. Note that it is not clear what Haudricourt would posit for the ‘year’, ‘egg’, and ‘long’ sets. Table 5-10 Comparison of solutions to “Gedney’s Puzzle” ‘year’ ‘egg’ ‘fire’ ‘long’ Li Sarawit Haudricourt *i *i: *ɤj *aj *ɛj *ia:j *CaCi *ej *ɤj Although it is extremely likely that some kind of “umlaut”-like sound changes were operating at some point in the history of these etyma, it cannot be the case that they occurred at the PT stage or later. First of all, this issue is closely tied to the canonical shape of the PT prosodic word. As discussed in §2.3.2, it is unlikely that the PT prosodic word was disyllabic. Haudricourt’s view is thus ruled out by the sesquisyllabic theory that we have earlier argued for. Moreover, for the umlaut view to work, we must assume two separate one-step reductions of disyllables to monosyllables in SWT and CT on one hand, and in NT on the other. That is, *CaCi must have been reduced directly to *Cai in SWT/CT and to *Ci: in NT without any intermediate step. This is because such an intermediate step would have to involve the same sort of complex vowels that Li and Sarawit posit. Another piece of evidence against Haudricourt’s view is that it falsely predicts only two types of languages. Languages pertaining to the first type are those that 203 merged the ‘long’ set with the ‘year’ set, and merged the ‘fire’ set with the ‘egg’ set. SWT and CT are generally said to belong to this type. The second type consists of those varieties that merged the ‘long’ set with the ‘egg’ set, and merged the ‘fire’ set with the ‘year’ set. NT is viewed as belonging to this type. In contrast, Li’s and Sarawit’s accounts allow at least eight types of languages as schematized in Table 5-11. Table 5-11 Types of languages allowed by Li’s and Sarawit’s account ‘year’ ‘fire’ ‘long’ ‘egg’ Type 1 Type 2 Type 3 Type 4 Type 5 Type 6 Type 7 Type 8 Tai varieties of type 1 do not show any mergers and have four different reflexes for the four sets. Varieties belonging to type 2 merged the ‘year’ set with the ‘fire’ set, and the ‘egg’ set with the ‘long’ set. Yay represents this type of language. Dialects belonging to type 3 show two reflexes for the four sets. They merged the ‘year’ set with the ‘long’ set, and the ‘egg’ set with the ‘fire’ set. Siamese and Lungchow both belong to this type. Types 4 and 5 merged three of the four sets together. Type 4 has the same reflex for the ‘fire’, the ‘long’, and the ‘egg’ sets while type 5 merged the ‘year’, ‘fire’, and ‘long’. Similar to type 3, types 6 and 7 have three reflexes for the four sets. Type 6 merged the ‘year’, and the ‘fire’ sets, but kept the ‘egg’ and the ‘long’ sets apart. In contrast, type 7 has one reflex for the ‘long’ and the ‘egg’ sets, but keeps the ‘year’ and the ‘fire’ sets distinct. Lastly, languages belonging to type 8 show only one reflex for the four sets. This last type is not informative as it can be arrived at from different paths. In short, Li’s and Sarawit’s accounts allow for 204 patterns of mergers that differ from the two found in Siamese, Lungchow on one hand, and Yay on the other. Even though it was formerly believed that only languages belonging to type 2 and type 3 exist, languages belonging to other types are in fact attested in southern Guangxi as shown in Table 5-12. Table 5-12 Tai varieties that do not belong to type 2 or type 3 ‘year’ ‘gall’ ‘fat’ ‘fire’ ‘dry field’ ‘chicken louse’ ‘long’ ‘excrement’ ‘egg’ ‘tree’ ‘intestine’ Shangsi po:yA1 bo:yA1 po:yA2 fo:yA2 lo:yC2 lo:yA2 lo:yA2 kʰo:yC1 lajB1 fajC2 ɬajC1 Chongzuo pɤjA1 bjɤjA1 pɤjA2 fɤjA2 lɤjC2 lɤjA2 Ningming pɤjA1 wɤjA1 pɤjA2 fɤjA2 lɤjC2 lajA2 -v Wuming pi:A1 bi:A1 pi:A2 fojA2 rojC2 rojA2 lɤjA2 kʰɤjC1 lɤjA2 kʰɤjC1 rajA2 xajC2 lajB1 fajC2 ɬajC1 lajB1 fajC2 ɬajC1 rajB1 fajC2 θajC1 type 5 type 7 Shangsi, Chongzuo, and Ningming are all type-5 languages because they merged the ‘year’, ‘fire’, and ‘long’ sets but kept ‘egg’ set apart. Wuming may belong to type 7. It seems to have merged the ‘long’ and the ‘egg’ sets but left the ‘year’ and ‘fire’ sets distinct. However, some etyma in the ‘year’ set show /oj/ rather than the expected /i/, e.g. ‘older sibling’ and ‘bladder’. Note that the Ningming form for ‘chicken louse’ unexpectedly shows /aj/. Based on the actual attestation of these types, 205 I adopt Li’s and Sarawit’s general position and propose that the four sets go back to four different rime categories as schematized in Figure 5-1. ‘fire’ *aj *ɤj *i: *vɤjA ‘fire’ *aɰ *ɤɰ *ɯ: *cɤɰA ‘breath’ *aw *ɤw *u: *plɤwB ‘empty’ ‘long’ *aj *ɯj *i: *rɯjA ‘long’ *aw *ɯw *u: *pɯwB ‘crab’ Figure 5-1 Proposed solution for “Gedney’s Puzzle” In this proposal, each of the four sets of etyma goes back to a distinct rime category. The etyma in the ‘year’ set had either *i:, *ɯ:, or *u: while members of the ‘egg’ set go back to PT *aj, *aɰ, *aw. Etyma that belong to the ‘fire’ set are reconstructed with either *ɤj, *ɤɰ, and *ɤw. These three rimes merged with *i:, *ɯ:, and *u: in NT languages as well as Shangsi, Ningming, and Chongzuo but merged with *aj, *aɰ, and *aw in SWT, most CT languages, and Wuming. In contrast, etyma belonging to the ‘long’ set are reconstructed with *ɯj or *ɯw. These two rimes merged with *i: and *u: respectively in SWT and CT, including the three type-5 dialects but merged with *aj, and *aw in NT varieties including Wuming. In addition, this proposal neatly accounts for the fact that etyma showing *ɯ: in SWT/CT but *aɰ in NT are not attested. If they existed, these items would belong to the ‘long’ set. According to the proposal, they would have to be reconstructed with 206 PT *ɯɰ, which would not be distinguishable from PT *ɯ:. We see, then, not only that this proposal accounts for “Gedney’s Puzzle” by positing relatively simple rime categories, it also correctly predicts a gap in the data. Often lumped together with the “Gedney’s Puzzle” items are some etyma that show irregular reflexes. Treating those items as belonging to the “Gedney’s Puzzle” paradigm gives an impression that modern dialects freely choose between the monophthongal or diphthongal forms of the so-called “Gedney’s Puzzle” items (Gedney 1989b: 60-61). In fact, these aberrant cases should be treated separately as exceptions, as they are cases of function words or words that occur with high frequency. It is largely unpredictable which vocalic variant a given dialect has, as illustrated in Table 5-13. Table 5-13 Etyma with irregular reflexes usually discussed under “Gedney’s Puzzle” ‘1s pron.’ ‘2s pron.’ ‘good’ ‘to give’ ‘to go’ PT Siamese *ku:A/*kawA ku:A1 *mɯŋA/*maɰA mɯŋA2 *ɗɤjA di:A1-v *haɰC haɰC1 *pajA pajA1 Tai Nüa kawA1 maɰA2 di:A1 -v hɯ:C1 -v pajA1 Lungchow kawA1 mɯŋA2 dajA1 hɯC1 -v pajA1 Shangsi kawA1 maŋA2 do:yA1 do:yA1 po:yA1 -v Wuming kawA1 mɯŋA2 dajA1 haɰC1 pajA1 The first-person singular pronoun seems to have been two separate forms in PT, *ku:A and *kawA. Similarly, its second-person counterpart seem to have also been two separate forms *mɯŋA, and *maɰA65 (Refer to Strecker (1984) for a detailed 65 The distinction between *ku:A and *mɯŋA, on one hand, and *kawA and *maɰA on the other, may have been prosodically conditioned allomorphy. In particular, *ku:A 207 study of PT personal pronouns). The etymon ‘good’ is generally rendered in modern Tai dialects with the reflex of PT *ɤj, except in SWT dialects, represented here by Siamese and Tai Nüa, where it shows the reflex of *ɯj. Similarly, the verb ‘to give’ seems to go back to PT *aɰ, but shows reflexes of *ɯ: in the CT dialects of Ningming, Longzhou, and Daxin, as well as some SWT dialects, e.g. Tai Yuan. Lastly, the verb ‘to go’ clearly had *aj in PT but unexpectedly has the reflex *i: in Shangsi, and Chongzuo. These unexpected cases must be treated as irregular and must be kept separate from the more regular “Gedney’s Puzzle” sets. Also note that the etyma in this set are either function word, or lexical items with a high frequency of occurrence. 5.5 Enriching the codas The conventional reconstruction of PT holds that only eight consonants can function as codas: *-p, *-t, *-k, *-m, *-n, *-ŋ, *-j, and *-w (Li 1977: 58-59). This characterization is consistent with the syllable typology of modern Tai languages. However, the discovery of the displaced NT language Saek (Gedney 1989c; Haudricourt 1958; 1960; 1963a) has opened up the possibility that the coda system of PT may have been richer than we originally assumed. In this section, I argue that, in addition to final *-l preserved in Saek, final *-c and possibly *-ɲ must also have existed in PT. Moreover, I show that these three consonants play a vital role in the development of the vowels. One of the intriguing characteristics of Saek is its final /-l/, unattested elsewhere in Tai. Li (1977: 58-59) does not reconstruct final *-l for PT but notes that some etyma that he reconstructs with *-n may in fact got back to *-l. The reason for and *mɯŋA were possibly the unstressed forms while *kawA and *maɰA were stressed. 208 his reservation is lack of Saek data. The situation changed dramatically after Gedney’s extensive study of this language (Hudak 1993). By examining the distribution of /-l/ in Saek, he argues convincingly that Saek /-l/ is a case of retention (Gedney 1993). First of all, he points out that most Saek words with final /-l/ are native Tai words. It is impossible to find a possible conditioning environment that changes final *-n to /-l/ in a great number of words but left it unaffected in many others. More importantly, he shows that final /-l/ in Saek is amply attested in rimes directly inherited from PT but is absent from rimes that were introduced later. Therefore, it is safe to conclude that PT had rimes ending in *-l. Comparison between Saek and Laha, a Kra language, also supports Gedney’s view that PT had final */-l/ (Ostapirat 1995). Many forms found in both Saek and Laha agree in having /-l/ as shown in Table 5-14. In addition to the cognates cited by Ostapirat (1995), I also cite ‘body louse’ as another Kra-Dai etymon with final /-l/. The Laha and Proto-Kra forms are taken from Ostapirat (2000). The etymon ‘grandchild’ is problematic because the Saek form has /-n/ instead of the expected /-l/. The final *-l may have become /-n/ in Saek due to dissimilation from the onset /l-/. Table 5-14 Shared etyma with final /-l/ in Saek and Kra ‘husked rice’ ‘shake’ ‘body louse’ ‘grandchild’ ‘to snore’ Saek sa:lA1 sɛlB1 mlɛlA2 la:nA1 trɛlA1 Laha sa:lA1 salB1 mdalA2 kla:lA1 kalA1 Proto-Kra *salA *səlB *m-drəlA *klalA 209 Although Saek is the only Tai language that has final /-l/, traces of PT*-l can also be found in the vowel correspondences among its sister languages. Most revealing are the data sets given in Table 5-15. The PT mid front vowel /e/ stays front in SWT and CT e.g. *cetD > Siamese /cetD1/ and Lungchow /citD1/. In contrast, it normally became *a in NT varieties, e.g. PT *cetD > Yay /satD1/ and Saek /cɛtD1/. Except for a few cases, Saek has changed earlier short *a to /ɛ/. The unexpected /-n/ in ‘grandchild' is perhaps result of dissimilation or dialect mixing. Table 5-15 Etyma with PT */en/ and */-el/ ‘seven’ ‘mushroom’ ‘great-grandchild’66 ‘body louse’ ‘to become’67 ‘civet cat’ Siamese cetD1 hetD1 le:nA1-t lenA2 penA1 -i henA1 Lungchow citD1 vitD2 -i minA2 pinA1 -i hinA1 Qinzhou citD1 lytD1 linC1 manA2 panA2 Yay satD1 ratD1 nanA2 panA2 ɲanA1 Saek cɛtD1 (hetD1) (le:nA1) mlɛlA phalA2 -v ɲɛlA1 PT *cetD *ʰrwetD *ʰlenC *mlelA *belA *ʰɲelA In Qinzhou, however, there is a split between ‘seven’, ‘mushroom’, and ‘greatgrandchild’ on the one hand, and ‘body louse’, ‘to become’, and ‘civet cat’ on the other. The vowels in the first set of etyma have been raised but still retain their front quality. The /y/ in ‘mushroom’ results from a coalescence of *-w- and *i. In contrast, the vowels in the second set have not been raised but were retracted and lowered to /a/. As all these items have final /-l/ in Saek, the backing of PT *e in these words should 66 The original tone for this etymon is most likely *C but it became *A in SWT due to contamination from *ʰla:nA ‘grandchild’. The vowel in the Siamese form has been secondarily lengthened. 67 This form is found in /panA2 jeŋB2/ ‘how’. 210 be considered to have been conditioned by PT *-l. Note the contrast between ‘greatgrandchild’ and ‘body louse’. Although they are both reflected with /-n/ in Qinzhou, the vowels are different. Therefore, these two sets of data show that final /-l/ caused PT *e to split into /i/ and /a/ in Qinzhou. Also note that the Saek forms for ‘mushroom’ and ‘great-grandchild’ are borrowings from Siamese or Lao. In addition to final *-l, evidence from Saek as well as from non-Tai languages points to the existence of final *-c in PT. Even though final palatals are common in SEA, especially among Mon-Khmer languages, they are not attested in any modern Tai language. However, Gedney (1989c: 961) notes that some of the Saek etyma that show a change from *at to /ɛt/ sometimes show a further change to /ɛk/. These etyma are shown in Table 5-16. I propose that etyma belonging to this set in fact go back to final *-c. Table 5-16 Etyma with final */-c/ ‘fish scale’ ‘ant’ ‘grain’ ‘to cut’ ‘to dust off’ Siamese kletD1 motD2 metD2 tatD1 patD1 Lungchow kitD1 mɤtD2 tatD1 patD1 Yay cipD1 -f68 matD2 natD2 tatD1 patD1 Saek tlɛkD1 mɛkD2 mlɛkD2 tɛkD1 pɛkD1 PT *klecD1 *mɤcD2 *mlecD2 *tacD1 *pacD1 One may argue that the final /-k/ in Saek is an innovation as it is preceded in all these examples by a front vowel /ɛ/. However, this cannot be the case for two 68 There exist many cases of sporadic /-p/ instead of the expected /-t/, e.g. Southern Shan /hepDS1/ and Ningming /ɬapDS1/ from *ʰrwitD ‘mushroom’, and Lungming /kip/, Western Nung /cip/, and Wuming /klip/ from *klecD ‘fish scales’ 211 reasons. First, the front quality of the vowel in Saek cannot be said to have been the condition for a change from *-t to /-k/ because many words with the rime /ɛt/ do exist, e.g. cɛtD1 ‘seven’ and /rɛt D1/ ‘fart’. Remember that Saek /ɛ/ before coronal codas corresponds regularly to /a/ in other NT varieties. Second, the Siamese and Lungchow forms show clearly that this mysterious coda consonant was preceded by many different vowels including *ɤ, *e, and *a. These vowels do not form a natural class that could potentially condition a change from *-t to /-k/. Therefore, the “aberrant” final /-k/ in Saek should be considered evidence for PT *-c. Evidence from outside of Tai also supports the hypothesis that PT had final *-c. First of all, Be, one of Tai’s closest kin, shows unique reflexes in items given in Table 5-17. PT *-t corresponds regularly to Be /-t/ as in ‘fart’, and PT *-k regularly to Be /k/, as ‘bird’. However, the etyma that show the /-t/~/-k/ correspondence in Tai have final /-ʔ/ in Be. The fact that Be shows this three-way contrast supports the hypothesis that PT had a distinct final *-c. This correspondence set has been noted by Ostapirat (2005: 119), who believes that PT *-c was inherited from Proto-Kra-Dai. Since it has been established that PT allows palatal consonants in the coda, one would also expect to find the palatal nasal occurring in coda position. Although the reconstruction of PT *-c is unequivocal, there is rather little evidence for final *-ɲ. The only potential case I have identified so far is ‘to eat’, which is reflected as /kinA1/ in all SWT varieties but as /kɯnA1/ in NT dialects like Wuming and Yay. We can speculate that the PT form for ‘to eat’ was *kɯɲA but the vowel was fronted so that the PSWT form for this etyma was *kinA. Therefore, I tentaitively hypothesize that PT had both *-c and *-ɲ. 212 Table 5-17 Final *-c in Tai and Be ‘fart’ ‘flea’ ‘bedbug’ ‘bird’ ‘deep’ ‘mucus’ ‘fish scale’ ‘ant’ ‘grain’ ‘to cut’ ‘to dust off’ Siamese Saek totD1 rɛtD1 matD1 matD1 rɯətD2 ruətD2 nokD2 nɔkD2 lɯkD2 lakD2 mu:kD2 mu:kD2 kletD1 tlɛkD1 motD2 mɛkD2 metD2 mlɛkD2 tatD1 tɛkD1 patD1 pɛkD1 Be ɗutD1 matD1 liətD2 nokD2 lakD2 mukD2 liʔD1 muʔD2 moʔD2 ɓaʔD1 5.6 Reconstructions of PT rimes Etyma from Siamese (SI), Lungchow (LC), and Yay (Y), are provided in each section to represent the three clusters of the modern varieties. Remember that this is not to say that they represent three genealogical subgroups of Tai (See §1.2 and §7.3 for discussion). 5.6.1 Rimes with high vowels The PT high vowels *i, *i:, *ɯ, *ɯ:, *u, and *u: are relatively stable, especially when compared to the mid vowels. Except in a few environments, these high vowels normally remain high in all languages. There seems to have been no restriction on the codas that particular vowels could co-occur with. However, short vowels were not allowed in open syllables due to the minimality requirement discussed in §2.2.2.1. 213 5.6.1.1 Rimes with *i and *i: PT high vowels *i and *i: are usually kept as high vowels in modern dialects. In open rimes, the long vowel *i: retains its original quality in most dialects, including Siamese, Lao, Black Tai, Yay, and Wuming. However, in some CT and NT dialects such as Lungming, Debao, Yongnan, Shangsi, Hengxian, and Longsheng, PT *i: became diphthongized in open syllables. An interesting case of a diphthongized reflex is /o:y/ in Shangsi, which is quite different from /ɤj/, /ej/, /ɯj/, or /uj/ in other dialects. PT *i:# corresponds to *i# in Li’s reconstruction (1977: 262-263). Examples of PT *i:# are given in Table 3-50. Table 5-18 Etyma with PT *i:# Gloss ‘soot’ ‘four’ ‘year’ ‘to have’ PT *ʰmi:C *si:B *pi:A *mi:A SI mi:C1 si:B1 pi:A1 mi:A2 LC mi:C1 ɬi:B1 pi:A1 mi:A2 Y mi:C1 θiC1 piA1 miA2 Others Lungming /mejC1/ Lungming /sejC1/ Lungming /pejA1/ Lungming /mejA2/ In closed syllables, both *i and *i: can occur. In some dialects such as Siamese, Lao, and Saek, the original quantity contrast is retained. In some other dialects, the contrast has been lost but the tonal reflexes on checked syllables provide clues to the original length of the etyma. For example, Southern Shan only has a length contrast for /a/ and /a:/ but shows different tonal reflexes for checked syllables with PT *i: and *i, e.g. /lip55/ from *ɗipD ‘raw’ vs. /nip21/ *ʰni:pD ‘to pinch’ (see §6.5.1 for discussion). Li reconstructs *iC (Li 1977: 261-262) and *iəC (Li 1977: 263-264) for 214 *iC and *i:C in the current reconstruction, respectively. Table 5-19 provides examples of PT *iC and *i:C. Table 5-19 Etyma with PT *iC and *i:C Gloss ‘to taste’ ‘short’ ‘pangolin’ ‘raw’ ‘foot’ ‘to climb’ ‘knife’ ‘to pinch’ PT *ɟimA *tinB *lilB *C̥ .dipD *ti:nA *pi:nA *mi:tD *ni:pD SI cʰimA2 dipDS1 ti:nA1 pi:nA1 mi:tDL2 ni:pDL1 LC cimA2 tinB1 linB1 dipDS1 mitDL2 nipDL1 Y simA2 tinB1 linB1 dipDS1 tinA1 pinA1 mitDS2 nipDS1 Others Black Tai /tinB1/ Black Tai /linB1/ Lungming /tinA1/ Lungming /pinA1/ Siamese normally preserves the original quantity contrast. In other words, PT *i is reflected as /i/ while PT *i: is reflected as /i:/ in Siamese. However, it has undergone a secondary shortening. In particular, PT *i: became shortened in Siamese words with the rising tone69, e.g. /hinA1/ ‘stone’ rather than */hi:nA1/, and /jiŋA1/ rather than */ji:ŋA1/. This is in sharp contrast with Lungchow and Yay, which both lost the distinction between the two high vowels without trace. Note that *i: in checked syllables regularly shortened in Yay as well as most other NT dialects, e.g. *mi:tD > /mitDS2/ instead of /mitDL2/. 69 This tone goes back to PT *A on syllables with PT fricatives, and voiceless sonorants in word-initial position. 215 5.6.1.2 Rimes with *ɯ and *ɯ: The high back unrounded vowels *ɯ and *ɯ: are the least stable among the PT high vowels. Like their front counterparts, only long *ɯ: was allowed in open rimes. They are preserved as back unrounded vowels in most dialects including Siamese, Black Tai, Lue, Lungchow, Western Nung, Daxin, Yay, Tianlin, Du’an, Saek, etc. Similarly to *i:#, this PT back vowel shows a diphthongal reflex /ɤɰ/ in such dialects as Lungming, Debao, Hengxian, and Longsheng. However, many dialects show special developments. For example, Qinzhou, Yongnan, Fusui, and Shangsi merged PT *ɯ:# with *i:# and now have /uj/, /ej/, /ɯj/ and /o:y/ respectively as reflex for both. PT *ɯ:# corresponds to *ɯ# in Li’s reconstruction (1977: 262263). Examples of PT *ɯ:# are given in Table 5-20. Table 5-20 Etyma with PT *ɯ:# Gloss PT ‘hand’70 *mwɯ:A ‘day’ *mwɯ:C ‘day after tomorrow’71 *C̬ .rɯ:nA ‘to buy’ *z.ɟɯ:C ‘writing, book’ *sɯ:A SI LC mɯ:A2 mɯ:A2 mɯ:C2 mɯ:C2 rɯ:nA2 lɯ:A2 sɯ:C2 ɬɯ:C2 sɯ:A1 ɬɯ:A1 Y fɯŋA2 fɯC2 rɯ:nA2 sɯC2 θɯA1 Others In closed rimes, both the short *ɯ and the long *ɯ: were allowed. Similarly to their front counterparts, the original quantity contrast is retained only in some varieties. However, unlike the other high vowels, the quality of PT *ɯ and *ɯ: have been 70 The final /-ŋ/ in Yay and other NT dialects for this etymon is aberrant. 71 Refer to footnote 53 for discussion on the aberrant /-n/ in the Siamese form. 216 affected by the codas. For example, they are lowered to /ɤ/ in many dialects including Lungchow, Leiping, and Lungchow. This development is seen most clearly when the rimes had an alveolar coda. PT *ɯ:n and *ɯ:l are reflected as /ɤn/ in Lungchow. For these two sets of etyma, Li (1977: 265) reconstructs *ɯən. In Siamese and other SWT languages, short *ɯn lowered to *ɤn (> PSWT *on) but long *ɯ:n did not. Li reconstructs (Li 1977: 274) *wɯn for this PT short rime72. Table 5-21 Etyma with PT *ɯn, *ɯ:n, and *ɯ:l respectively Gloss PT ‘sky’ *ɓɯnA ‘rain’ *q.wɯnA ‘human, person’ *ɢwɯnA ‘night’ *ɣɯ:nA ‘other’ *ʔɯ:nB ‘firewood’ *wɯ:lA ‘to slide’ *mlɯ:lB SI bonA1 fonA1 kʰonA2 kʰɯ:nA2 ʔɯ:nB1 fɯ:nA2 lɯ:nB2 LC pʰɤnA1 kɤnA2 kɤnA2 ʔɤnB1 fɤnA2 Y bunA1 hunA1 hunA2 hɯnA2 funA2 Others Po-ai /mɯnA1/ Po-ai /hɯnA1/ Po-ai /hɯnA2/ Wuming /ʔɯnB1/ Saek /mlɯ:lB2/ This development of rimes with alveolar codas above is in contrast to rimes with final labial consonants. PT *ɯm and *ɯ:m became *um and *u:m in many dialects including Longzhou, Qinzhou, Guangnan Nung, Yay and Wuming. The rounding of *ɯm and *ɯ:m must have been due to the influence of *-m. Li 72 Coincidentally, both ‘human’ and ‘rain’ had medial *-w- in PT. However, this medial *-w- is not the same as the *w in Li’s *wɯn. The etymon ‘sky’ clearly illustrates this fact as it did not have *-w- in PT, i.e. PT *ɓɯnA > ɓɤnA > PSWT ˀbonA > Siamese /bonA1/. The rounding in the Siamese forms was not caused by the *-w-. Rather, it is a result of a regular change from *ɤ to PSWT *o, cf. PT *sɤnC ‘heal’ > Siamese /sonC1/. 217 reconstructs *ɯum (1977: 265) and *wom (1977: 273-274) for these two PT rimes. The development of *ɯŋ and *ɯ:ŋ is in turn quite different from that of *ɯm and *ɯ:m. In contrast, PT *ɯ and *ɯ: before velar codas are usually lowered to /ɤ/ or /a/ in SWT, CT, and NT dialects. Both *ɯŋ and *ɯ:ŋ are reflected as /ɯŋ/ in Siamese, but as /aŋ/ and /ɯ:ŋ/ in Ningming. Since the only etymon with PT *ɯ:ŋ is not found in NT, it is not clear what its reflex is. Li (1977: 264) reconstructs *ɯŋ for the short rime but did not give a reconstruction for its long counterpart. Etyma in Table 5-22 show the development of PT *ɯm, *ɯ:m, *ɯŋ, and *ɯ:ŋ. These rimes are illustrated by ‘wind’, ‘to forget’, ‘to steam’, and ‘one’ respectively. Table 5-22 Etyma with PT *ɯm, *ɯ:m, *ɯŋ, and *ɯ:ŋ respectively Gloss ‘wind’ ‘to forget’ ‘to steam’ ‘one’ PT *C̬ .lɯmA *lɯ:mA *ʰnɯŋC *ʰnɯ:ŋB SI LC Y lomA2 lumA2 rumA2 lɯ:mA2 lumA2 lumA2 nɯŋC1 naŋC1 nɯŋB1 nɤŋB2 -t Others Ningming /naŋC1/ Ningming / nɯ:ŋB1/ In addition to the effect of places of articulation discussed above, manners of articulation of the coda also came into play in the development of the high back unrounded vowel. For example, *ɯt became *ut in Po-ai, in contrast to *ɯn above which remains /ɯn/. Another example is the development of PT *ɯ: before the velar stop *-k. This rime became /u:k/ in Siamese, and /uk/ in Lungchow, but remains unrounded in Po-ai and Yay /ɯk/. Notice the regular shortening of high vowel *ɯ: preceding stop codas in NT, parallel to *i: and *u: (see §5.6.1.1and §5.6.1.3). The correspondents of PT *ɯ:t, and *ɯ:k in Li’s system are *ɯut (1977: 265), and *ɰuk 218 (1977: 268) respectively. Notice the contrast between *ɯu and *ɰu in Li’s system. In Table 5-23, ‘distance from thumb to fingertip’ represents *ɯ:p, ‘insipid’ represents *ɯ:t, and ‘child’ and ‘to come into contact’ represent *ɯ:k. Also, note that *ɯ:p rounded to /up/ in Yay because of the labial coda, in parallel to *ɯm and *ɯ:m above. Table 5-23 Etyma with PT *ɯ:p,*ɯ:t, and *ɯ:k respectively Gloss ‘distance from thumb to fingertip’ ‘insipid’ ‘child’ ‘to come in contact’ PT *ɣɯ:pD SI kʰɯ:pDL2 *cɯ:tD cɯ:tDL1 *lɯ:kD lu:kDL2 *C̬ .tɯ:kD tʰu:kDL1 LC lukDL2 tʰukDS1 Y hupDS2 lɯkDS2 tɯkDS2 Others Po-ai /ɕutDS1/ The coda is not the only component of the syllable that can affect the development of the vowel; the medial glide could also play a crucial role. One very important change of PT *ɯ: that is conditioned by the medial *-w-. In particular, PT *ɯ:# fused with the medial *-w- that preceded it, resulting in special developments as shown in Table 5-24. The sequence *-wɯ: became simplified to *u: in CT and SWT but became a diphthong *ɯə in most NT dialects. The NT dialect of Lingyue seems to be the only dialect that still retains the medial *-w-, cf. /lwaA2/ ‘ear’ and /ŋwaA2/ ‘snake’. Li (1977: 268) posits *ɰu# for this set of data. 219 Table 5-24 Etyma with PT *wɯ:# Gloss ‘ear’ ‘snake PT *k.rwɯ:A *ŋwɯ:A SI hu:A1 ŋu:A2 LC hu:A1 ŋu:A2 Yay rɯəA2 -i ŋɯəA2 Others Lingyue /lwaA2/ Lingyue /ŋwaA2/ Also refer to §5.4 for the reconstruction of PT *ɯw and *ɯj. 5.6.1.3 Rimes with *u and *u: PT high vowels *u and *u: are usually kept as high vowels in modern dialects. In open rimes, the long vowel *u: retains its original quality in most dialects, including Siamese, Lao, Black Tai, Yay, and Wuming. However, some CT and NT dialects such as Lungming, Debao, Yongnan, Shangsi, Hengxian, and Longsheng, PT *u: became diphthongized to /ɤw/, /ow/, or /aw/ in open syllables. PT *u:# corresponds to *u# in Li’s reconstruction (Li 1977: 266-267). Examples of PT *u:# are given in Table 5-25. Table 5-25 Etyma with PT *u:# Gloss ‘pig’ ‘rat’ ‘to be at’ ‘door’ ‘pair’ PT *ʰmu:A *ʰnu:A *ˀju:B *tu:A *gu:B SI mu:A1 nu:A1 ju:B1 tu:A1 kʰu:B2 LC mu:A1 nu:A1 ju:B1 tu:A1 ku:B2 Y mu:A1 nu:A1 ɲu:B1 tu:A1 ku:B2 Others Lungming /mowA1/ Lungming /nowA1/ Lungming /jowB1/ Lungming /towA1/ Lungming /kowB2/ 220 In closed syllables, both *u and *u: can occur. In some dialects such as Siamese, Lao, and Saek, the original quantity contrast is retained. In some other dialects, the contrast has been lost but the tonal reflexes on checked syllables provide a clue to the original length of the etyma. For example, Southern Shan only has a length contrast for /a/ and /a:/ but shows different tonal reflexes for checked syllables with PT *u and *u:, e.g. /tup55/ from *dupD ‘to beat’ vs. /lup21/ *C̬ .lu:pD ‘to stroke’ (see §5.3 for discussion). Li reconstructs *uC (Li 1977: 266) and *uəC (Li 1977: 267) for *uC and *u:C in the current reconstruction respectively. Table 5-26 provides examples of PT *uC and *u:C. Table 5-26 Etyma with PT *uC and *u:C respectively Gloss ‘to cook’ ‘to be entangled’ ‘puppet’ ‘to beat’ ‘to carry in the arm’ ‘to lead’ ‘mucus’ ‘to stroke’ PT *ʰruŋA *ʰɲuŋB *hunB *dupD *ʔu:mC *cu:ŋA *mu:kD *C̬ .lu:pD SI LC Y huŋA1 huŋA1 ruŋA1 juŋB2 juŋB2 ɲuŋB2 hunB1 hunB1 tʰupDS2 tupDS2 ʔumC1 ʔumC1 ʔumC1 cu:ŋA1 cu:ŋA1 mu:kDL2 mu:kDL2 mukDS2 lu:pDL2 rupDS2 Others W. Nung /hunB1/ W. Nung /tupDS2/ W. Nung /lupDS2/ Siamese normally preserves the original quantity contrast. PT *u is reflected as /u/ while PT *u: is reflected as /u:/. However, the Siamese form for ‘to carry in the arm’ underwent a secondary shortening, possibly conditioned by tone. Notice the regular shortening of *u: to /u/ before stop codas in NT, illustrated by Yay /mukDS2/ 221 ‘mucus’ and /rupDS2/ ‘to stroke’. This shortening also occurred with long *i:C (see §5.6.1.1) and *ɯ:C (see §5.6.1.2). Note that *uk went through special developments in SWT and NT as illustrated by the data in Table 5-27. Table 5-27 Etyma with PT *uk Gloss ‘bone’ ‘loom’ ‘tomorrow’73 PT *C̥ .dukD *trukD *ɟm̩ .rukD SI du:kDL1 hu:kDL1 pʰru:ŋDL2 LC Y dukDL1 dokDL1 hukDL1 rokDL1 pjukDL2 sokDL2 Others Lungming /nokDS1/ Lungming /lokDS1/ Lungming /cokDL2/ Corresponding to *uok in Li’s system (1977: 267-268), PT *uk was lengthened to *u:k in all SWT and many CT dialects including Lungchow, Leiping, Ningming, etc. In all NT varieties including Yay, *uk became lengthened and lowered to *o:k. The modern forms in Siamese, Lungchow, and Yay all show DL tones, indicating an earlier long vowel. There are some CT dialects that retained the original short vowel. For example, Lungming and Western Nung have DS tones in these etyma, pointing to an earlier short vowel. 5.6.2 Rimes with mid vowels PT had altogether three pairs of short and long mid vowels: *e, *e:, *ɤ, *ɤ:, *o, and *o:. Like their high vowel counterparts, short vowels were not allowed in open syllables due to the minimality requirement discussed in §2.2.2.1. However, unlike PT 73 The Siamese form shows /-ŋ/ instead of the expected /-k/ due to assimilation to /ni:C2/, cf. expression /pʰru:ŋDL2 ni:C2/ ‘tomorrow’ in present-day Siamese. 222 high vowels, the long *e: and *o: regularly lowered to *ɛ: and *ɔ: in a great number of SWT and NT dialects. 5.6.2.1 Rimes with *e and *e: PT had two mid front vowels *e and *e:, which contrasted with each other in quantity. PT *e:# was lowered to *ɛ:# in SWT and some CT dialects. Examples of languages that underwent the lowering include Siamese, Lao, Southern Shan, Black Tai, Lue, Sapa, Bao Yen, Cao Bang, and the NT language Saek. The PT *e: is generally preserved as a mid vowel in the remaining CT and NT dialects. Like their its vowel counterparts, only long *e: can occur in open rimes due to the minimality constraint on the prosodic word as discussed in §2.2.2.1. PT *e:# corresponds to *ɛ# in Li’s reconstruction (1977: 273-274). Examples of PT *e:# are given in Table 5-28. Table 5-28 Etyma with PT *e:# Gloss ‘old (of living beings)’ ‘to untie’ ‘mother’ ‘raft’ ‘fish net’ PT *ke:B *ke:C *me:B *be:A *kre:A SI kɛ:B1 kɛ:C1 mɛ:B2 pʰɛ:A2 hɛ:A1 LC ke:B1 ke:C1 me:B2 pe:A2 he:A1 Y ceB1 meB2 peA2 reA1 Others Wuming /ke:C1/ In closed syllables, both *e and *e: can occur. In SWT dialects including Siamese, Lao, Black Tai, etc., PT *e in closed syllables is generally retained as a mid vowel. However, it is raised to /i/ before nasal codas in some dialects such as Lue and White Tai. Among CT varieties, it is raised to /i/ in most dialects including 223 Lungchow, Lungming, Leiping, Shangsi, and Qinzhou. It is, however, preserved as /e/ in a few varieties including Daxin and Guangnan Nung. In all NT dialects including Yay, Saek, and Wuning, PT *eC is regularly raised to /iC/. Li (1977: 269-270) reconstructs *je for this set of data. Table 5-29 illustrates the development of PT *eC. Table 5-29 Etyma with PT *eC Gloss ‘full’ ‘needle’ ‘child (young person)’ ‘to pick up’ ‘fingernail’ PT *k.temA *qemA *ɗekD *kepD *C̬ .lepD SI temA1 kʰemA1 dekDS1 kepDS1 lepDS2 LC timA1 kʰimA1 dikDS1 kipDS1 lipDS2 Y rimA1 cimA1 cipDS1 ritDS2 -f Others However, there is a special development shared by most NT languages including Saek, Yay, Po-ai, Yongbei, Hechi, Rong’an, Guigang, etc. In these varieties, PT *e when followed by an alveolar coda became /a/ so that PT *et, *en, and *el are now reflected as /at/, /an/ and /an/ respectively. Table 5-30 provides example of *et, *en, and *el. Notice that the Lungchow forms evidence the general raising of *eC to *iC discussed above. 224 Table 5-30 Etyma with PT *et, *en, and *el respectively Gloss ‘seven’ ‘mushroom’ ‘porcupine’ ‘great-grandchild’ ‘body louse’ ‘civet cat’ PT *cetD *ʰrwetD *ʰmenC *lenC *mlelA *ʰɲelA SI LC cetDS1 citDS1 hetDS1 vitDS1 menC1 minC1 le:nA1 -t lenA2 minA2 henA1 hinA1 Y ɕatDS1 ratDS1 nanA2 ɲanA1 Others Saek /manC1/ Po-ai /lanC1/ Saek /mlɛlA2/ Saek /ɲɛlA1/ In contrast to *e, PT *e: in closed rimes follow the same course of development as *e: in open rimes. That is, it is reflected as /ɛ:/ in SWT and some CT varieties but preserved as /e:/ in most dialects. Examples of PT *e:C are given in Table 5-31. Li (1977: 273-274) reconstructs *ɛC for this set of data. Table 5-31 Etyma with PT *e:C Gloss ‘cheek’ ‘drought’ ‘strong, strength’ ‘eight’ ‘to burst’ PT *ke:mC *C̬ .le:ŋC *re:ŋA *pe:tD *p.re:kD SI kɛ:mC1 lɛ:ŋC2 rɛ:ŋA2 pɛ:tDL1 tɛ:kDL1 LC ke:mC1 le:ŋC2 ɬe:ŋA2 pe:tDL1 pʰe:kDL1 Y cemC1 reŋC2 reŋA2 petDL1 pekDL1 Others 225 5.6.2.2 Rimes with *ɤ and *ɤ: PT mid back unrounded vowels *ɤ and *ɤ: show a wide range of reflexes in modern dialects. Like the other mid vowels, *ɤ and *ɤ: only contrasted in closed syllables. In a number of CT dialects including Western Nung and Jingxi, as well as most NT dialects including Yay, Saek, Qiubei, Shanglin, Guigang etc., *ɤC lowered to /aC/. In all SWT varities including Siamese, Lao, Lue, Black Tai, and White Tai, PT *ɤC is generally reflected as /oC/. A number of CT and NT varieties such as Shangsi, Yongnan, Laibin, and Yishan also underwent the rounding of *ɤ. In contrast, in a number of CT dialects including Lungchow, Leiping, Lungming, Debao, and Chongzuo, to cite a few, the short *ɤC remains /ɤC/, except before labial codas, where they became rounded and raised to /up/ or /um/. Table 5-32 gives examples of etyma that go back to PT *ɤC. The current interpretation of the data largely agrees with Sarawit (1973) and Wang (1966) but differs from Li (1977: 271-272), who posits *oC for this set of etyma. Table 5-32 Etyma with PT *ɤC Gloss ‘ant’ ‘heel’ ‘fart’ ‘chest’ ‘frog’ ‘sour’ PT *mɤcD *sɤnC *k.tɤtD *ʔɤkD *kɤpD *sɤmC SI motDS2 sonC1 totDS1 ʔokDS1 kopDS1 somC1 LC mɤtDS2 ɬɤnC1 tɤtDS1 ʔɤkDS1 kupDS1 ɬumC1 Y matDS2 θanC1 ratDS1 ʔakDS1 kapDS1 θamC1 Others 226 The reflex of PT long *ɤ: in closed rime is in sharp contrast with *ɤ. Following Sarawit (1973) and Wang (1966), *ɤ:C have two main reflexes. In NT and a few CT dialects, *ɤ:C simply lowered to /a:C/ in parallel with *ɤC. Languages that show this development include Yay, Saek, Guigang, Longsheng, Qiubei, Qinzhou, Yongnan etc. The second group of dialects includes SWT and CT varieties such as Siamese, Dehong, Lao, Lue, Longzhou, Longming, Dabao, Daxin, etc. In these dialects, PT *ɤ:C diphthongized to /ɯəC/ in contrast to the other long mid vowels *e: and *o: which simply lowered to /ɛ/ and /ɔ/ (see §5.6.2.1 and §5.6.2.3). Table 5-33 provides examples of etyma with PT *ɤ:C. Li (1977: 282) posits *ɯa for this set of data. Note that Lungchow /ɤ:C/ comes from earlier *ɯəC (< PT *ɯaC and *ɤ:C). Table 5-33 Etyma with PT *ɤ:C Gloss ‘house’ ‘rope’ ‘boil (intr.)’ PT rɤ:nA ɟɤ:kD ɗɤ:tD SI rɯənA2 cʰɯəkDL2 dɯətDL1 LC ɬɤ:nA2 cɤ:kDL2 dɤ:tDL1 Y ra:kA2 sa:kDL2 da:tDL1 Others In open syllables, the minimality requirement allowed long *ɤ: but bars short *ɤ from occurring. PT *ɤ:# is raised, merging to *ɯ: in SWT and CT dialects including Siamese, Lao, Lungchow, Lungming, Debao, Jingxi, etc. In all NT dialects and a few CT varieties including Yongnan, Long’an, Fusui, Shangsi, and Chongzuo, it became rounded and merged with PT *o:#. Qinzhou is special in that it is the only dialect that simply lowered *ɤ:# to /a:/. A handful examples of *ɤ:# are available, as given in Table 5-34. Note that ‘meat’ shows a peculiar development in SWT and CT 227 due to its onset, i.e. *n.mɤ:C > PSWT *nɯəC > Siamese /nɯəC2/ vs. *n.mɤ:C > *mlɤ:C > Yay /noC2/. Table 5-34 Etyma with PT *ɤ:# and *ɤ:l Gloss ‘straight’ ‘name’ ‘meat’ ‘new’ ‘to put’ PT *zɤ:B *ɟɤ:B *n.mɤ:C *mɤ:lB *s.cɤ:lB SI sɯ:B2 cʰɯ:B2 nɯəC2 majB1 sajB1 LC ɬɯ:B2 cɯ:B2 nɯ:C2 maɰB1 Y θoB2 soB2 noC2 moB1 soB1 Others Qinzhou /ɬa:B2/ Qinzhou /na:C2/ Qinzhou /moB1/ Also included in Table 5-34 are etyma with *ɤ:l, which shows a different development from *ɤ: with other codas. In SWT and most CT, PT *ɤ:l became *ɤɰ earlier on, merging with original *ɤɰ. It is now reflected as /aj/ in Siamese but /aɰ/ in Lungchow (see §5.4). In NT and a few CT dialects, PT *ɤ:l merged with *ɤ:# so that it is now reflected as /o/ in Yay. Also refer to §5.4 for the reconstruction of PT *ɤj, *ɤɰ, and *ɤw. 5.6.2.3 Rimes with *o and *o: The PT mid back rounded vowels *o and *o: went through different paths of development. In parallel with *e:#, PT long *o:# was lowered to /ɔ:/ in SWT and some CT dialects. Examples of languages that underwent the lowering include Siamese, Lao, Southern Shan, Black Tai, Lue, Sapa, Bao Yen, Cao Bang, and the NT language Saek. PT *o:# is still generally preserved as a mid vowel in the remaining CT and NT dialects. Like its high vowel counterparts, only the long *o: can occur in open rimes 228 due to the minimality requirement on PT prosodic word as discussed in §2.2.2.1. PT *o:# corresponds to *ɔ# in Li’s reconstruction (1977: 277-278). Examples of PT *o:# are given in Table 5-28. Table 5-35 Etyma with PT *o:# Gloss ‘stump’ ‘well, spring’ ‘shaman’ ‘bee’ ‘neck’ PT *ko:A *ɓo:B *ʰmo:A *to:B *ɣo:A SI kɔ:A1 bɔ:B1 mɔ:A1 tɔ:B1 kʰɔ:A2 LC ko:A1 bo:B1 mo:A1 to:B1 ko:A2 Y koA1 boB1 moA1 toB1 hoA2 Others In closed syllables, both *o and *o: can occur. The *o however did not develop exactly like its front counterpart *e. In SWT and CT dialects including Siamese, Lao, Black Tai, Leiping, Lungming, etc., PT *o in closed syllables is generally retained as a mid vowel. However, it is raised to /u/ before nasal codas in some dialects such as Lue and White Tai. In contrast, a few CT varieties raised *oC to /uC/ across the board. Some examples of these dialects are Lungchow, Western Nung, and Qinzhou. In parallel with *eC, in most NT dialects including Yay, Tianlin, Donglan, and Wuming, PT *oC is regularly raised to /uC/. Li (1977: 272-273) reconstructs *wo for this set of data. Table 5-36 illustrates the development of PT *oC. 229 Table 5-36 Etyma with PT *oC Gloss ‘to boil (tr.)’ ‘buttock ‘to itch’ ‘to shrink’ ‘crooked’ PT *tomC *konC *ɣomA *ʰrotD *gotD SI tomC1 konC1 LC tumC1 kun C1 hotDS1 kʰotDS2 kutDS2 Y tumC1 humA2 rutDS1 Others S. Shan /kʰomA2/ However, there is a special development shared by most NT languages including Yay, Po-ai, Yongbei, Hechi, Rong’an, Guigang, etc. In these varieties, PT *o when followed by a velar coda did not raise to *u so that PT *ok, and *oŋ remain /ok/ and /oŋ/. Table 5-37 provides examples of *o followed by velar codas. This set of data is reconstructed with *ok, and *oŋ in Li’s system (1977: 271-272). Table 5-37 Etyma with PT *ok, and *oŋ respectively Gloss ‘bird’ ‘to drop’ ‘dry land’ ‘to take down’ ‘forest’ ‘lost’ ‘knee’ ‘to blow’ PT *C̬ .nokDS *tokD *ɓokD *ploŋA *ɗoŋA *ʰloŋA *χowB *m.powB SI nokDS2 tokDS1 bokDS1 ploŋA1 doŋA1 loŋA1 kʰawB1 pawB1 LC nukDS2 tukDS1 bukDS1 pjuŋA1 duŋA1 luŋA1 kʰawB1 pawB1 Y rokDS2 tokDS1 bokDS1 pjoŋA1 doŋA1 loŋA1 hoB1 poB2 Others 230 Note a special development of *ow. In SWT and CT dialects, PT *ow became *aw merging with the original PT *aw (see §5.4). The change in vowel quality is probably a dissimilation in contact with final *-w. In NT, the rime *-ow became *o:# merging with original PT *o:# discussed above. Li (1977: 291) also posits *ow for this set of data. In contrast to *o, PT *o: in closed rimes followed the same course of development as *o: in open rimes. That is, it is reflected as /ɔ:/ in SWT and some CT varieties but preserved as /o:/ in most dialects. Examples of PT *o:C are given in Table 5-38. This set of data is reconstructed with *ɔC in Li’s system (1977: 277-278, 287-288). Table 5-38 Etyma with PT *o:C Gloss ‘to exit’ ‘blind’ ‘fragrant’ ‘drum’ ‘sugarcane’ PT *ʔo:kD *ɓo:tD *ho:mA *klo:ŋA *ʔo:jC SI ʔɔ:kDL1 bɔ:tDL1 hɔ:mA1 klɔ:ŋA1 ʔɔ:jC1 LC ʔo:kDL1 bo:tDL1 ho:mA1 kjo:ŋA1 ʔo:jC1 Y ʔokDL1 botDL1 homA1 coŋA1 ʔojC1 Others 5.6.3 Rimes with low vowel *a and *a: The low vowels *a and *a: are the most stable of all the PT vowels. Like other vowel pairs, only the long *a: was permitted in open syllables due to the minimality 231 requirement. PT *a:# is generally retained as long /a:/ 74 . Li (1977: 275-276) reconstructs *a for this PT long vowel. Examples of PT *a:# are given in Table 5-39. Table 5-39 Etyma with PT *a:# Gloss ‘to kill’ ‘dog’ ‘fish’ ‘to scold’ ‘to open mouth’ PT *qa:C *ʰma:A *pla:A *ɗa:B *ʔa:C SI kʰa:C1 ma:A1 pla:A1 da:B1 ʔa:C1 LC kʰa:C1 ma:A1 pja:A1 da:B1 ʔa:C1 Y ka:C1 ma:A1 pja:A1 da:B1 ʔa:C1 Others The length contrast between PT *a and *a: in closed rimes is usually retained even in varieties that have lost it elsewhere. For example, high and mid vowels in Wuming do not contrast for length but low vowels /a/ and /a:/ do, cf. /plakDS1/ from PT *prakD ‘vegetable’ vs. /pla:kDL1/ from *pra:kD ‘forehead’. Moreover, varieties that have lost the length contrast completely still have different reflexes for *a and *a:. For example, Bouyei dialects substituted the original length in low vowels by a quality contrast so that it has /ɑ/ for *a and /a/ for *a:, cf. /pjɑkDS1/ from PT *prakD ‘vegetable’ vs. /pjakDL1/ from *pra:kD ‘forehead’. Table 5-39 provides examples of PT *aC and *a:C. These two rimes correspond to *əC (Li 1977: 270-271) and *aC (Li 1977: 275-276) in Li’s system. 74 Since /a/ and /a:/ never contrast in open syllables, data from Chinese sources usually analyze [a:#] as /a/. I systematically re-transcribe it to /a:/. 232 Table 5-40 Etyma with PT *aC and *a:C respectively Gloss ‘to take’ ‘black’ ‘to spin’ ‘to steal’ ‘to take a bite’ ‘father’s younger brother’ ‘village’ ‘handle’ ‘to drag’ ‘mustard green’ PT *ʔawA *ɗamA *panB *ʰlakD *katD *ʔa:wA *ɓa:nC *ɗa:mA *C̬ .la:kD *ka:tD SI ʔawA1 damA1 panB1 lakDL2 katDS1 ʔa:wA1 ba:nC1 da:mA1 la:kDL2 ka:tDL1 LC ʔawA1 damA1 panB1 lakDL2 katDS1 ʔa:wA1 ba:nC1 da:mA1 la:kDL2 ka:tDL1 Y ʔawA1 damA1 panB1 rakDL2 katDS1 ʔa:wA1 ba:nC1 da:mA1 ra:kDL2 ka:tDL1 Others The above sets of data provide a clear illustration of the contrast between *aC and *a:C. Not only are the reflexes still distinguished by vowel length, their tonal reflexes also attest the original vowel length contrast. For example, Siamese /lakDS2/ from PT *C̬ .lakD ‘to steal’ has DS2 tone but /la:kDL2/ from PT *C̬ .la:kD ‘to drag’ shows DL1. Also, note a special development of *aC in etyma with medial *-w-. In many CT and NT dialects including Debao, Jingxi, Yay, Wuming, Yongbei and Saek, the medial *-w- and the vowel *a usually fused together to become /ɔ/ or /o/, e.g. Saek /ɣɔnA2/ from *ɣwanA ‘smoke’. This special rounding apparently did not occur with *a:, cf. Saek /va:nA2/ from *xwa:nA ‘axe’. 233 5.6.4 Rimes with diphthongs Among the five diphthongs *iə, *ɯə, *uə, *ɤɰ, and *aɰ in PT, the three rising diphthongs generally pattern together. Generally, they are still preserved in some SWT dialects including Siamese, Lao, and Black Tai but have been monophthongized in all other dialects. The results of the monophthongization are mid vowels /e/, /ɤ/, and /o/ in Phu Thai, White Tai, Southern Shan, Dehong, and Lue, but high vowels /i/, /ɯ/, and /u/ in Lungming, Leiping, Debao, Yongnan, Chongzuo, Wuming, Yishan, Qiubei etc. Some varieties such as Yay, Lingyue, and Tiandong however only monophthongized PT *uə but kept *iə, and *ɯə in tact. In Lungchow, *ɯə gives /ɤ:/ (< *ɯ:, see §5.6.1.2) in closed syllables but /ɯ:/ in open syllables. Table 5-41 gives examples of etyma with PT *iə, *ɯə, *uə. These PT diphthong corresponds to *ie, (Li 1977: 280281) *ɯe (Li 1977: 281-282), and *ue (Li 1977: 283-284) in Li’s system. Table 5-41 Etyma with PT *iə, *ɯə, and *uə respectively Gloss ‘to lick’ ‘sticky ‘disgusted’ ‘moon’ ‘blood’ ‘deaf’ ‘to scald’ ‘coconut grub’ PT SI *C̬ .liəA liəA2 *ʰniəwA niəwA1 *ɓɯəB bɯəB1 *ɓlɯənA dɯənA1 *lɯətD lɯətDL2 *nuəkD nuəkDL1 *luəkD luəkDL2 *ɗuəŋC duəŋC1 LC li:A2 bɯ:B1 bɤ:nA1 lɤ:tDL2 nu:kDL1 lu:kDL2 Y riəA2 niəwA1 dɯənA1 lɯətDL2 nu:kDL1 lu:kDL2 duŋC1 Others 234 Note a special development widely found in all the three groups. PT *iəw simplified to *ɛ:w in many dialects including Lungchow, Leiping, Yay, Wuming etc. For example, Lungchow has /de:wA1/ for PT *ɗiəwA ‘single’ rather than */di:wA1/ indicating that the change *iəw > *e:w occurred before the change from *iə to /i:/. Siamese still retains the original rime, cf. Siamese /diəwA1/. Table 5-42 gives examples of PT *iəw. Table 5-42 Etyma with PT *iəw Gloss ‘single, only one’ ‘withered’ ‘green’ PT *ɗiəwA *ʰriəwB *xiəwA SI diəwA1 hiəwB1 kʰiəwA1 LC de:wA1 he:wB1 kʰe:wA1 Y dewA1 rewB1 hewA1 Others The falling diphthong *aɰ developed very differently from the rising ones. In a number of languages across the three dialect clusters, *aɰ merged with *aj so that it is now reflected as /aj/ in Siamese, Lao, Tai Yuan, Yongnan, Qinzhou, Tiandong, Hechi, etc.; /ɔj/ in Debao and Jingxi, or /o:y/ in Shangsi and Chongzuo. In contrast, the PT *aɰ simplified to /ɤ/ in some NT dialects including Saek, Liujiang, Yishan, and Rong’an. Of course, there are varieties that still retain the diphthong as /aɰ/, including Black Tai, White Tai, Lungchow, Lungming, Daxin, Yanshan Nung, Yay, Wuming, Shanglin, Laibin, etc. Table 5-43 presents some etyma that go back to PT *aɰ. This PT diphthong corresponds to Li’s *ɤɰ (1977: 288-289). 235 Table 5-43 Etyma with PT *aɰ Gloss ‘leaf’ ‘clear, clean’ ‘gizzard’ ‘near’ ‘daughter-in-law’ PT *ɓaɰA *saɰA *p.taɰA *k.raɰC *baɰC SI bajA1 sajA1 tajA1 klajC1 pʰajC2 LC baɰA1 saɰA1 taɰA1 kʰraɰC1 Y baɰA1 saɰA1 taɰA1 caɰC1 paɰC2 Others The diphthong *ɤɰ will not be discussed here because it has already been discussed in relation to the so-called “Gedney’s Puzzle” (see §5.4). Note that PT *ɤɰ and *aɰ are written as if they were sequences of vowel plus glide to reflect the fact that they pattern with *ɤj, *ɤw, *aj, and *aw with regards to “Gedney’s Puzzle”. However, phonologically they are clearly complex vowels. First of all, they do not show length contrast. If they had been sequences of vowel plus glide, we would expect *ɤ:ɰ and *a:ɰ to have existed. Moreover, *ɰ never occurs with other vowels, i.e. sequences like *eɰ, *oɰ, *iɰ and *uɰ did not exist. If *ɰ had functioned as a final consonant, it should have been found in combination with other vowels as well. 5.7 Summary In this chapter I have proposed a system of PT rimes that differ markedly from the conventional reconstruction (Li 1977). First, I have shown that PT had a fullfledged vowel length contrast which interacted closely with syllable structure. Second, I have accounted for the so-called “Gedney’s puzzle” by positing four sets of rimes that do not add any complexity to the system. Lastly, I proposed that final *-l, *-c, and perhaps *-ɲ also existed in PT in addition to those assumed in the conventional 236 reconstruction. The evidence for these codas comes from Saek as well as less aberrant varieties of Thai. Unlike previous proposals, the current system takes seriously the effect of medial and coda consonants on the development of the vowels. These improvements eliminate the need to posit improbably fine contrasts or unrealistically complex vowels in the PT phonology. Not only does this proposal account better for the whole range of data, it is also simpler and more symmetrical. 237 CHAPTER 6 PROTO-TAI TONES 6.1 Introduction A reconstruction of a proto-language is an attempt to present a concrete picture of that linguistic system. With this in mind, the reconstruction of PT cannot be complete without addressing the poorly understood issue of PT tone shapes. This statement is at odds with the general impression that the tonal system of PT has already been firmly established. It is true that a great amount of research has clearly shown that PT had four tonal categories, conventionally labeled *A, *B, *C, and *D (Gedney 1972a; Li 1977; Nishida 1954; Wulff 1934). However, the reconstructed system is still abstract, as the concrete characteristics of each tone are still unknown. One major problem is the lack of an adequate methodology for tonal reconstruction. Recently, some attempts to reconstruct the phonetic characteristics of proto-tone systems have been made, e.g. Proto-Southwestern Tai (Brown 1965), Old Black Tai (L-Thongkum 2002), Proto-Shan (Morey 2004), and Proto-Northern Min (Handel 2003), but most of them are confined to a group of closely-related dialects whose tonal systems are almost identical, both in terms of contrasts and realizations. This methodological inadequacy in turn stems from poor theoretical and empirical understanding of tone change in general. Faced with this challenge, this chapter approaches the question not only from a comparative perspective but also from the standpoint of systematic similarities between tonal categories across varieties. In this chapter, I first review the literature on PT tones highlighting those findings that will be crucial in the reconstruction of the PT tone shapes. Then, I present a brief theoretical discussion of the mechanisms that underlie tonal changes. Then, I propose a concrete reconstruction of PT tonal categories with respect to pitch 238 height, pitch contour, vowel duration, and voice quality. The evidence for such a system comes mainly from modern reflexes of the tonal categories. I claim that PT belongs to the “pitch-plus-voice quality” type of tone language. In other words, both pitch and voice quality were crucial to the identity of PT tones. Lastly, I present an inventory of PT etyma classified according to their reconstructed tones. 6.2 3+1: The PT tonal system One of the most important advancements in the field of Comparative Tai is the successful reconstruction of tonal categories in PT. Because the methodology and theory behind the reconstruction has been discussed in great detail in great many fora (Edmondson & Solnit 1997; Gedney 1972a; Li 1977), in this section I only briefly highlight those aspects of the PT tonal system that are crucial to the understanding of PT phonology as well as review the earlier reconstruction of PT tones. The PT tonal system can be characterized as a “three-plus-one” tonal system. Most, if not all, researchers working on the history of Tai languages agree that the proto-language had three contrastive tones, conventionally labeled *A, *B, and *C. However, the contrast among these tonal categories only existed in “non-checked” syllables, or syllables that ended in a vowel or a sonorant consonant. In addition to these three categories, a fourth category *D can be said to have existed in checked syllables, or syllables that have a stop coda. Since the *D tone is the only tone that occurs in checked syllables, one may say that checked syllables neutralize the contrast observed in non-checked syllables. Phonologically, what distinguished *D from *A, *B, and *C was its syllable structure. Despite this lack of tonal contrast, it was probably not the case that *D did not have characteristic tonal properties associated with it. Rather, it is likely that its phonetic characteristics resembled those of one of 239 the three tones on non-checked syllables. However, for convenience I follow standard practice in referring to as if it was a separate tonal category. Clearly, it is not an accident that the *ABCD system reconstructed for PT resembles those systems found in Chinese (Baxter 1992; Karlgren 1964; Li 1980; Pulleyblank 1991) and other language groups in SEA (Haudricourt 1961), e.g. Hmong-Mien (Chang 1975) and Vietnamese (Haudricourt 1954b), etc. It is not trivial that all these groups have historically been in intense contact with Chinese. The connections between the PT and Chinese tonal systems are especially striking. Not only are the two systems structurally identical, their tones also show robust correspondences. Specifically, vocabulary shared between Tai and Chinese permits us to establish connections between Chinese tones and Tai tones (Li 1976; Manomaivibool 1975; Wulff 1934) as schematized in Table 6-1. The Middle Chinese forms are from Pulleyblank (1991). The correspondences between Tai and Chinese tones will provide useful information for the reconstruction of PT tones (see §6.5). Table 6-1 Correspondences between Tai and Chinese tones PT *A *so:ŋA ‘two’ *bwi:A ‘fat’ Middle Chinese 雙 ṣåŋ Even 肥 bjwei 平 *B *ha:nB ‘goose’ *ɣe:ŋB ‘shin’ 雁 ŋanC 脛 ɣieŋC Departing 去 *C *ha:C ‘five’ *ma:C ‘horse’ 五 ŋuoB 馬 maB Rising 上 *D *pe:t ‘eight’ 八 păt *ŋɯək ‘crocodile’ 鱷 ŋâk Entering 入 240 Table 6-1 illustrates the tonal correspondence between PT and MC. Etyma that had *A in PT regularly showed Even tone in MC, as exemplified by ‘two’ and ‘fat’. Similarly, etyma that show *B in PT had Departing tone in MC. Lastly, etyma that had *C in PT had Entering tone in MC. This systematic correspondence between Tai and Chinese is useful in the reconstruction of PT tones as well as in the dating of Chinese borrowings into Tai. However, this is not to say that all Chinese loans in PT came from a single stage of Chinese. As shown convincingly by Manomaivibool (1975; 1976), many Chinese etyma in Tai must have been borrowed prior to the MC period. Moreover, it is also true that there are many cases where the generalization does not hold. For example, *kajB ‘chicken’ had *B in PT but its Chinese source 雞 jī (< MC kiei) had Even tone. These cases have been analyzed by Manomaivibool (1975). Nevertheless, why Old Chinese and Middle Chinese loans behave similarly in this respect requires an in-depth study. Even though the tonal systems among modern Tai varieties differ greatly, they can always be shown to have developed from this *ABCD tonal system (Gedney 1972a). However, none of the modern dialects keeps the three-plus-one system intact. This is because daughter languages have all undergone tonal splits conditioned by the laryngeal features of the onsets. In all Tai languages, the four original tones have split into at least two series, those reflecting original voiceless and those reflecting original voiced onsets. This kind of voicing-based tonal split across *A, *B, *C, and *D tones may be referred to as “register split”. In addition to voicing, other laryngeal features that condition tonal splits in Tai languages include aspiration, glottalization, and frication. In checked syllables, vowel length may introduce a tonal split in the *D category. It is crucial to note that it is the laryngeal features of the onset at the time of splits rather than the original PT onset features that were the determining factors in tonal splits. Presumably, an onset feature 241 at any period could have triggered a tonal split, but many of the features that did this were not present in the proto-language. For example, many SWT and CT dialects including Siamese, Southern Thai, Lungming, etc. have gone through tonal splits based on aspiration but this does not mean that PT had contrastive aspiration. Rather, the aspiration-based tonal splits indicate that by the time the splits occur those individual dialects had already developed a series of contrastive aspirated stops. If the tonal splits had to refer necessarily to the original PT laryngeal features, aspirationbased tonal split could not have arisen in any Tai dialects, as PT did not contrast for aspiration (see §3.5). To complicate the picture, tones that resulted from such splits may later merge, further obscuring the tonal correspondence among different dialects. The best tool for capturing the history of tonal splits and mergers in individual Tai varieties is generally known as the “tone-box” method (Gedney 1972a), briefly discussed in §1.4.3. Table 6-2 shows phonological factors that have been commonly found to have conditioned tonal splits and mergers in Tai dialects including voicing, aspiration, and glottalization. Because tonal reflexes provide clues to the phonation types of the onsets, tones in modern languages are referred to according to their etymological sources (see §1.4.3). For instance, using “1” and “2” to represent original voiced and voiceless onsets respectively, tone A2 refers to tone in a given daughter language that developed from PT *A tone conditioned by voiced onsets. Simarly, tone DS1 refers to tone in a given in dialect whose etymological source is tone *D with short vowel conditioned by voiceless onsets. 242 Table 6-2 Phonological factors commonly found to condition tonal splits and mergers (modified from Gedney 1972a) *A *B *C *DS *DL aspirated *f-, hm-, ph-,… 1 unaspirated *p-,… glottalized/implosive *ɓ-,… 2 voiced *v-, m-, b-,… Most modern varieties show a simple register split, i.e. bipartition across the four tonal categories based on voicing. Tones in modern dialects that were conditioned by earlier voiceless and glottalized onsets are referred to as A1, B1, C1, and D1 tones. Similarly, tones that were conditioned by earlier voiceless onsets are label A2, B2, C2, and D2 tones. The tonal systems of White Tai and Western Nung, given in Table 6-3 and Table 6-4, respectively, represent dialects with this kind of basic split. The difference between the two is that the latter further split *D according to vowel length, but the former did not. The *D tone with short vowels is labeled *DS while the *D tone with long vowels is usually referred to as *DL. Because tones in checked syllables are analyzed as belonging to the same categories as those in non-checked syllables, these types of dialects are expected to show six tones altogether. However, the Southern Shan dialect described by Cushing (Hudak 2008) given in Table 6-5 later merged B2 and C1 to yield five tones instead of six. 243 Table 6-3 Pattern of tonal splits and mergers in White Tai aspirated unaspirated glottalized voiced *A 1 *B 2 *C 3 *DS *DL 22 4 5655 Table 6-4 Pattern of tonal splits and mergers in Western Nung aspirated unaspirated glottalized voiced *A 1 *B 2 *C 3 *DS *DL 62 4 5645 244 Table 6-5 Pattern of tonal splits and mergers in Southern Shan aspirated unaspirated glottalized voiced *A 1 *B 2 *C 3 *DS *DL 42 4 3553 One fact that is crucial to the reconstruction of the shapes of the PT tones is that *DL very often patterns with *B for the purposes of tonal mergers and splits (Chamberlain 1975; Gedney 1989a). In other words, the two categories often are sensitive to the same conditioning features in the onsets, and are often reflected as the same tones in modern dialects. An example is the case of Western Nung in Table 6-4, among others. In this variety, there was a simple register split in both *B and *DL. Moreover, the two categories agree in having tones 2 and 5 as reflexes of the first and second series, respectively. White Tai and Southern Shan also share this pattern of tone mergers with Western Nung. This strong connection between *B and *DL suggests that *D shared many phonetic properties with *B. Some varieties show tonal splits based on aspiration or glottalization in addition to the basic register split. The Southern Thai dialect of Surat Thani (LThongkum 1978) in Table 6-6 is a clear case of a variety that went through an aspiration-based tonal split. This variety underwent a three-way split across *A, *B and *C but later merged tones that developed from *A1 with tones that developed from *B1. In contrast, the NT dialect of Po-ai in Table 6-7 is an example of a variety that underwent a tonal split based on glottalization. In this case, the *A tone was split 245 into three tones—one from original voiceless onsets, one from original glottalized/implosive onsets, and one from voiced onsets. Table 6-6 Pattern of tonal splits and mergers in Southern Thai aspirated unaspirated glottalized voiced *A 1 3 *B *C *DS *DL 222 777 4 5665 Table 6-7 Pattern of tonal splits and mergers in Po-ai voiceless glottalized voiced *A 1 6 2 *B 5 6 *C 3 4 *DS *DL 25 3 6 Of particular interest are dialects that still retain the original PT voiced onsets despite having undergone register splits. These varieties include many CT dialects of northeastern Vietnam and southwestern Guangxi (Haudricourt 1960; Kosaka 1997; LThongkum 1997; Ross 1996) as illustrated by the dialect of Cao Bang in Table 6-8. Varieties of this type indicate that the binary register split and the devoicing of the voiced onsets, though closely connected, are two separate sound changes. The pattern of tone splits and mergers in Cao Bang is given Table 6-8, where each tone is also 246 annotated with Chao’s tone numbers (Chao 1930). Examples of the retained voicing contrast are provided. In this variety, the vowels are breathy after voiced onsets (Pittayaporn 2007d). Table 6-8 Pattern of tonal splits and mergers in Cao Bang aspirated unaspirated glottalized voiced *A *B *C *DS *DL 1 = [53] 3 = [43] 2 = [34] 5 3 5 = [21] 4 = [33] 6 = [25] 5 6 A1 = pi:1 ‘year’ < *pi:A B1 = tɤw3 ‘turtle’ < *tawB C1 = pa:2 ‘turtle’ < *pa:C A2 = bi:5 ‘fat’ < *bwi:A B2 = dɤw4 ‘ashes’ < *dawB C2 = bja:6 ‘knife’ < *ɟm.ra:C Moreover, Cao Bang and other dialects that retain voicing reveal how the basic binary register split and neutralization of phonation type in onsets occurred. In particular, they show that voicing of voiceless sonorants caused the original tones to split into two series (L-Thongkum 1997). Originally, the PT tonal categories were influenced by the phonation type of the onsets so that they had different phonetic realizations after voiced onsets and voiceless onsets. In the second step, these phonetic differences became phonologized, creating two categorically distinct pitch registers. These registers were redundant because they were still predictable from the phonation type of the onset. Subsequently, the voiceless sonorants became voiced, merging with the original voiced sonrants. It is this merger that caused the redundant registers to 247 become phonemic, resulting in a complete binary register split. The number of tones in each dialect consequently doubled. Cao Bang and other conservative dialects have only completed this step of the evolutionary path. The fourth step is the devoicing of the original voiced obstruents. Most Tai dialects have completed this step but the reflexes of the devoiced onsets could either be plain voiceless consonants or aspirated consonants. While Lue, Black Tai, Lungchow, Lungming, Yay, and most dialects show unaspirated reflexes, Siamese, Lao, Leiping, and Saek have aspirated reflexes. The interrelated processes of binary register split and neutralization of phonation type in onsets discussed is schematized in Figure 6-1. Stage I *ʰn-, *t*n- *d- *A, *B, *C, *D phonetic effect of phonation type on tonal realization Stage II *ʰn-, *t-, *n- *d- *A1, *B1, *C1, *D1 categorical but redundant pitch *A2, *B2, *C2, *D2 registers Stage III *t-, *A1, *B1, *C1, *D1 phonemic registers in sonorants *n- *d- *A2, *B2, *C2, *D2 Stage III *A1, *B1, *C1, *D1 pitch registers not predictble from *n- *t- *A2, *B2, *C2, *D2 onsets Further tonal splits and mergers Figure 6-1 Binary register split and neutralization of phonation types 248 Despite the fact that register split and devoicing are found through out Tai, it is not the case that these changes had already occurred before the diversification of PT. On the contrary, the binary register split and the devoicing took place relatively recently in the history of individual Tai dialects. Chamberlain (1975; 1991) dates the register split back to the PT period but such a position is untenable. The original distinction between voiced and voiceless onsets clearly survived until late in the history of modern dialects. For example, Ayutthaya inscriptions dated to the 15th century CE (Fine Arts Department 1986) still reflect the distinction between voiceless and voiced consonants (Gedney 1991; Pittayaporn 2007a). Moreover, feeding and bleeding relationships among changes indicate that the devoicing occurred only after Black Tai had changed *ɣ- to *g- (Pittayaporn 2007a). These kinds of evidence demonstrate beyond doubt that the interrelated process of register split and devoicing are relatively late events in the history of individual Tai languages. In talking about the realizations of tonal categories, one needs to seriously consider the role of voice quality in distinguishing among the tones. It is extremely common typologically for tones to be distinguished by voice quality in addition to pitch. A well-known case is Northern Vietnamese, in which three out of six tones exhibit salient glottal properties (Gedney 1989a; Maspéro 1912; Nguyễn & Edmondson 1998; Thompson 1965). The “curve” tone (hỏi) is falling-rising and usually creaky. The “broken” tone (ngã) starts with an abrupt fall followed by a sharp rise. The abrupt fall is accompanied by strong glottalization of vowel necleus. Similarly, the “drop” tone (nặng) is a falling tone which ends in a glottal constriction. In fact, some researchers have claimed that PT had contrastive phonation types (Sagart 1988), or lacked prosodic contrast (Gedney 1989d). In the remainder of this section, I review evidence for voice quality in *B and *C. 249 There is evidence that the *ABCD tonal system of PT may have made use of voice quality. Gedney (1989d) points out that in a great number of CT and SWT varieties tones that have developed from PT *C are characterized by glottalization. According to Gedney’s impressionistic observations, this glottalization is usually realized as extreme tension of the glottis throughout the articulation of the vowel, ending in a complete glottal constriction at the end of the syllable. Unfortunately, to date there have been no systematic instrumental studies of glottalization across dialects. Moreover, most dialect descriptions use the term glottalization very loosely without distinguishing between creakiness and final glottal constriction (see §6.5.5 for the distinction between creakiness and glottal constriction). However, from Gedney’s discussion it is safe to assume that the glottalization he proposes for *C was a final glottal constriction. Western Nung in Table 6-9 is a clear example of a dialect that has been described as having glottalized tones. Table 6-9 Western Nung tonal inventory A1 low rising [14] A2 mid-high level [44] B1 low falling [21] B2 mid falling [31] C1 mid-low level, glottalized [22ˀ] C2 high level, glottalized [55ˀ] In Western Nung, the two tones that developed from PT *C are realized as mid-low level and high level tones, both with clear glottalization. In contrast, the other tones, which come from PT *A and *B, do not show such voice quality. This strong tendency for tones that have developed from *C to be glottalized suggests that at least one of the PT tones may have been characterized by voice quality in addition to pitch. 250 Gedney (1989d) specifically proposes that PT *C tone had a final glottalization or final glottal stop at an earlier stage before the tonal splits occurred. Gedney’s proposal that PT *C was a glottalized tone is in contrast with Sagart’s hypothesis (1988) that *C was a creaky tone. Although Sagart recognizes that modern dialects usually show final glottalization for tones that developed from *C, he refers to it as “creakiness” and does not reconstruct *C with final glottalization. He accounts for glottalization in modern reflexes of *C by positing creaky voice quality. This choice of reconstruction is motivated by his view that it was *B that ended with a glottal stop. I will show shortly that this could not have been the case. As for PT *B, Gedney (1989d) speculates that it had a final *-h. First of all, he draws on the connection between *B and *DL to argue that *B may had been a final consonant like its *DL counterparts. Moreover, he shows that Indic words that end in /-h/, when borrowed into Siamese, lost the final fricative, and were assigned to tone *B. This leads him to speculate that *B tone in PT developed from final *-h. If this theory is correct, PT must be considered a completely non-tonal system. Specifically, the categories *A, *B, and *C would have to be characterized as plain syllables, syllables ending in *-h, and syllables ending in *-ʔ, respectively. Although this view that *B syllables may have had a final *-h at the PT level is not impossible, the tone assignment in Indic borrowings cannot be taken as evidence for this account. First of all, Gedney (1989d) himself points out that his account for *B is based only on data from a relatively recent layer of Siamese vocabulary. Second, he admits the possibility that the Indic words ending in /-h/ were assigned to the *B category because they were phonetically somehow similar to syllables ending in stops. In contrast to Gedney’s view of *B, Sagart (1988) hypothesizes that syllables with PT *B tone had a modal voice rime that ended with a glottal stop. He reasons that the fact that *B patterned with *D with respect to tonal mergers and splits indicates 251 that they ended with the same type of consonants, namely stops, cf. Vietnamese tonogenesis (Haudricourt 1954b). He also argues that positing final *-ʔ for *B accounts for Gedney’s Indic loanword data discussed above as well as the fact that Siamese words ending in /-ʔ/ belong to the DS category. Basically, he hypothesizes that Siamese incorporated Indic and Mon-Khmer borrowings with /-h/ by converting the final glottal fricative to a final glottal stop /-ʔ/. Consequently, those loanwords were naturally considered as belonging to the *D category. According to this explanation, the glottal stop was lost after a long vowel but is kept after a short vowel in Siamese. It is crucial to note that Sagart does not distinguish between creakiness on vowels and final glottalization. However, this explanation is unlikely for two reasons. Firstly, if PT *B was in fact a glottal stop /-ʔ/, we would expect to see a contrast between PT etyma with long vowel followed by a glottal stop, i.e. *CV:ʔ, and etyma with short vowel followed by a glottal stop, i.e. *CVʔ. However, there is no evidence for such contrast in PT. Sagart himself cites a personal communication with Ferlus, who correctly points out that many of the Siamese /CVʔB/ words are loanwords, especially from Khmer. Secondly, the final /-ʔ/ in Siamese is clearly epenthetic in nature and usually does not have an etymological source. Bennett (1995) shows that the final glottal stop in Siamese is phonologically inserted to satisfy word minimality which dictates that a stressed syllable must be heavy (see §2.2.2.1). In contrast to Gedney’s and Sagart’s interpretations, I view this set of Siamese borrowing data as a result of loanword adaptation. In particular, Indic etyma with final /-h/ are assigned to the *B category because they follow the same pattern as Indic words ending in other obstruents. Based on data in Gedney (1965), Indic words ending 252 in obstruents75, including the voiceless stops -p, -t, -c, and -k, and the fricatives -s, -ś, and -ṣ, were regularly assigned to the *D category when borrowed into Siamese, e.g. รถ /rotDS2/ ‘car’ from Sanskrit rath(a-), ทิศ /tʰitDS2/ ‘direction’ from diś, บาท /ba:tDL1/ from pād(a-), and โรค /ro:kDL2/ from rog(a-). Given that /-h/ is an obstruent, we would expect it to follow the same pattern. For instance, we expect */pʰa:hDL2/ ‘bearer’ from vāh(a-), */ʔutDS1sa:hDS1/ ‘to make an effort’ from utsāh(a-). However, the final /-h/ was in fact dropped in the process of loanword adaptation because final fricatives are not permissible in Siamese, leaving these etyma to end with an open syllable, cf. พา่ ห์ /pʰa:B2/ and อุตส่าห์ /ʔutDS1sa:B1/. Because Siamese B1 and B2 are the open-syllable counterparts of DL1 and DL2 respectively (refer to Table 6-12), these forms are considered to have tone B1 and B2 rather than DL1 and DL2. This explanation also accounts for Sagart’s observation that Siamese words ending in /-ʔ/ belong to the DS category. According to this scenario, the final /-h/ in loanwords with short vowels was dropped when incorporated into Siamese, which does not allow final fricatives. By dropping the /-h/, the loanwords are left with a short vowel without a coda. This kind of light syllable was banned by word minimality, triggering the /-ʔ/ epenthesis so as to preserve the original short vowel. For example, Khmer braḥ /preah/ ‘honorific prefix’ would have been incorporated into Siamese as *bra, but the phonology of PT automatically inserted /-ʔ/ so that the etymon was actually borrowed as *braʔ, which is now pronounced พระ /pʰraʔ/ in Siamese. This interpretation is supported by the fact that loanwords that do not have a final consonant in the source Indic language also have final /-ʔ/, e.g. Siamese อายุ /ʔa:A1 yuʔB2/ ‘age’ from āyu-, คณะ /kʰanaʔB2/ ‘group’ from gana- etc.76 Therefore, the tonal 75 Indic etyma were usually borrowed into Siamese with the final vowel truncated (Gedney 1965: 77). For example, Sanskrit เลข lekha- gives /le:kDL2/ without the final -a. 76 The fact that Khmer, through which Siamese borrowed most of its Indic vocabulary, has /ʔa:yuʔ/ for ‘age’, and /kʰanaʔ/ for ‘group’ indicates that the Indic sources did not 253 assignment of Indic loanwords in Siamese cannot be used as evidence for positing a laryngeal coda for PT *B as proposed by Gedney and Sagart. In sum, PT consisted of three contrastive tonal categories *A, *B and *C in non-checked syllables. Checked syllables, which had no tonal contrast, and are usually labeled *D for convenience, most likely had the same tone as *B. While it is unclear what the phonetic characteristics of those tonal categories were, modern reflexes of *C suggest that voice quality played an important role in the contrast between PT tones. In the following section, a discussion of mechanisms of tonal change is presented to provide the basis for the reconstruction of PT tones. 6.3 Directionality of tone change The typology of sound changes has long played a role in historical phonology, although this role is sometimes controversial (see (Hock 1991: 617-629)). The essential usefulness of an understanding of the typology of sound change is to clarify likely asymmetries in the directionality of change, e.g. [ki] > [ci] is more likely than [ci] > [ki]. Unfortunately, it is still a mystery how tonal contours change after tonal categories have been established, despite advances in research on tonogenesis and register split (e.g. Abramson & Erickson 1978; Haudricourt 1954b; Hombert, Ohala, & Ewan 1979; Matisoff 1973; Thurgood 2002b). In order to discuss the development of PT tones, it is necessary to have a framework that addresses the way tonal contours change over time. As a solution to this lack of research, I adopt the framework that I myself developed in Pittayaporn (2007b). Even though the framework is tentative, it is useful as a synthesis of earlier work on tone. end in a final consonant. Final consonants in Sanskrit are normally preserved in Khmer, e.g. /tɯt/ ‘direction’ from diś, and /kruəh/ ‘(mis)fortune’ from gṛha-. 254 Drawing on the advances in the phonetic study of lexical tone (e.g., Abramson 1978; Gandour 1994; Xu 1997; 2004), this framework attempts to establish a theory of tone change by positing mechanisms that predict the directions of tonal contour change. Following Ohala (2003), I assume that each phonological unit is realized with a great deal of orderly phonetic variation, as shown by fine-grained instrumental analyses. Studying this phonetic variation and speakers’ accommodation of ambiguity that it entails can provide a principled way of determining the likelihood of sound changes. In this framework, I claim that a change in tonal contour (i.e. change in the overall f0 trajectory) results from a reorganization of the tonal variants that occur in different environments (phonetic, phonological, stylistic, sociolinguistic, etc.). Drawing on synchronic studies of tonal variation, I propose three major mechanisms of tone change that make predictions about the directionality of tone change: 1) segment-tone interaction, 2) contextual variation, and 3) perceptual maximization. 6.3.1 Segment-tone interaction It is well-established that segments may cause variation in the f0 trajectory of tones (e.g. Hombert, Ohala, & Ewan 1979; Maddieson 1984a). I argue that the interaction between tones and segments may cause the tonal onset to shift. The most common effect of onset on f0 is the f0 lowering following voiced consonants, but other effects exist as well. Solnit and Kingston (1989) and Kingston (2004) discuss how higher f0 can develop after voiced onsets. In addition, Xu (2003) shows that the onsets of Mandarin tones are phonetically lowered after aspirated onsets. It seems that whether a given type of segment would raise or depress f0 does not follow automatically from the phonetics but depends on the phonology of the language (Kingston 2004; Kingston & Diehl 1994). 255 Under this view, any change in overall contour shape is an incidental consequence of the onset shift. Furthermore, the mechanism of segment-tone interaction is predicted not to affect the tone offsets. For example, for a case in which aspiration in the onset induces pitch lowering, a mid level tone would more likely yield a low rising tone than a low level one (see Figure 6-2a). Similarly, a mid level tone is more likely to become a high falling tone in cases where aspiration causes pitch raising. segment-tone interaction contour reduction peak sliding perceptual maximization (a) (b) (c) (d) Figure 6-2 Schematization of the mechanisms of tone change This claim is supported by the fact that the greatest effect is observed at the onset of the vowel. Note that the effects on the f0 values of vowels after initials of different laryngeal configurations may be significantly different long after the onset of the vowel (Hombert, Ohala, & Ewan 1979). 6.3.2 Contextual variation The phonetic realizations of lexical tones have been shown to vary across phonetic and phonological contexts (e.g. Gandour 1994; Xu 1997). I claim that the 256 fact that the underlying targets of lexical tones are often not fully reached due to such effects as contour reduction, peak delay, etc. (Xu 2004) causes “distorted” variants to be generalized and taken as the best approximation of the underlying representation. The intuition is that the frequency of contextually-affected contours is greater than that of “ideal” contours. 6.3.2.1 Contour reduction Dynamic tones are realized with reduced contours when the relevant syllables are in non-final positions. I claim that the mechanism of contour reduction only targets the contour shape and leaves the tonal onset intact. A change due to contour reduction can therefore be viewed as a change from a dynamic tone to a static tone without modification of the tonal onset. For example, a change from high falling tone to high level tone is more likely than a change in the opposite direction (see Figure 6-2b). This is supported by the fact that reduced variants show a flatter contour characterized by a smaller amount of f0 excursion, and a less extreme f0 value for the tonal offset. However, in these contours the tonal onset stays in the range expected for the onset of the full contour. Under this view, the less extreme f0 in the tonal offset is a result of the flattening of the contour shape. 6.3.2.2 Peak sliding I argue that it is more likely for the peak of a tonal contour to slide rightward than leftward. In such cases of peak sliding, the tonal onset should stay intact but the contour shape should change as a result of the rightward sliding of the peak. For example, a change from a falling tone to a convex tone is more likely than one in the opposite direction (see Figure 6-2c). 257 Xu (2001) finds that the f0 peak of a Mandarin syllable carrying a high or a rising tone may in fact occur in the following syllable. He argues that because sharp rises take a relatively long time to terminate, the f0 turning point may occur somewhat after the syllable boundary in certain environments and in fast speech. As f0 peaks tend to be delayed77 rather than early, tone peaks have a greater tendency to shift rightward than leftward. 6.3.3 Perceptual maximization As avoidance of perceptual confusions between phonologically distinct categories is usually hypothesized to be an important ingredient for successful communication (Flemming 2004), I claim that pitch height and pitch excursion may enhance each others to achieve maximal perceptual contrasts. Specifically, the greater the f0 excursion, the less likely a dynamic tone is to be perceived as static. For example, a change from a mid falling tone to a high falling tone is more likely than one in the opposite direction (see Figure 6-2d). This mechanism is based on Flemming’s (2004) argument that perceptual difficulties are not derived from properties of particular sounds but from constraints on categorization of speech sounds. In addition, Stevens et al. (1986) observe that contrastive sounds are often enhanced by redundant features that help listeners to perceive the distinctions. With respect to tones, Yip (2001) argues that the existence of dynamic tones allows a great number of contrasts without placing static tones tightly together in the perceptual space (also see Hombert 1977). As argued by t’Hart et al. (1990), the size of a pitch change must exceed a certain threshold to play a part in communicative functions. 77 While “peak sliding” refers to the diachronic displacement of the f0 peak, “peak delay” refers the synchronic phenomenon of late occurrence of the f0 peak. 258 While contour reduction tends to convert contour tones into static tones, perceptual maximization works within dynamic tones to heighten peak-trough contrasts. They need not have contradictory effects. If a dynamic tone has variants that differ in terms of the pitch value of the tone onset, this mechanism predicts that the variant that shows the greatest f0 excursion is likely to be generalized. The three mechanisms of tonal changed discussed in Pittayaporn (2007b) as summarized above will be the basis for the discussion of the development of PT tones in the following section. In particular, they will be referred to in discussing how modern tonal reflexes can be projected back to their respective three-plus-one systems. Having established the theoretical background for tonal changes, we now turn to the modern reflexes of the PT tonal categories. 6.4 Tonal reflexes of PT in daughter varieties Comparing modern reflexes of PT tones is a challenging task because of the magnitude of the variation among modern tone systems. Various changes that bring about splits and mergers of tonal categories have produced a very complex set of correspondences which makes strict application of the comparative method a daunting task. However, the picture becomes clearer when the data are organized from the point of view of the *ABCD system. The introduction of internal reconstruction helps reduce the number of correspondences, facilitating the reconstruction of the prototonal system. First, we arrive at the tonal system of the patterns of tonal splits and mergers of the proto-language in question. Then, we apply internal reconstruction to individual varieties to discover propertiess that they inherited from each of the prototones. These steps in effect reduce modern tone systems to *ABCD systems, whose tonal categories can be more conveniently compared. 259 In the following discussion, I will focus on the structural characteristics of each system that are helpful in arriving at an internal reconstruction. I will discuss the phonetic details of the change only when necessary. For our purposes, only seven tone systems are compared. The particular varieties are chosen because 1) their first-series tones and second-series tones can be straightforwardly collapsed into the original three-plus-one system, and 2) they represent different geographical regions of the Taispeaking area. These varieties include Siamese, Black Tai, Leiping, Cao Bang, Wuming, Liujiang, and Lianshan. Because the CT dialect of Cao Bang in northern Vietnam, spoken the center at the Tai-speaking territory, still preserves the original voiced onsets, and its first-series tones and their corresponding second-series tones still have similar tone shapes, its tonal system (given again in Table 6-10) is the first to be examined. Table 6-10 Pattern of tonal splits and mergers in Cao Bang78 aspirated unaspirated glottalized voiced *A *B *C *DS *DL 1 = [53] 3 = [43] 2 = [34] 5 3 5 = [21] 4 = [33] 6 = [25] 5 6 The main mechanism that brought about the observed tone system in Cao Bang is segment-tone interaction (see §6.3.1). The two tones that developed from *A both 78 This analysis is based on my own fieldwork. It differs from Hoang (1997)’s analysis in which the glottalized onsets patterned with the voiceless onsets in the *A column. 260 have falling contours, clearly pointing back to a high falling tone. Assuming that in Cao Bang’s register split voiced consonants depressed the onset of the tones, tone A2 (= tone 5) can be derived by lowering the onset of tone *A1 (= tone 1). Interestingly, the glottalized/implosive onsets pattern with the voiced onsets in this column, as in many other dialects. However, the phonetic motivation for this patterning is irrelevant. Similarly, the two tones that come from *B indicate an earlier mid falling tone. Lowering the onset of B1 (= tone 3) straightforwardly gave B2 (= tone 4). Notice that for *B the implosives patterned with voiceless initials. In contrast to *A and *B, Cao Bang *C must have a rising tone. Lowering of C1 (= tone 2) also straightforwardly gives C2 (= tone 6). Therefore, the original Cao Bang tonal system consisted of a high falling, a mid falling, and a mid rising tone. Note the connection between *B1 and *DL1. The next variety to be examined is the CT dialect of Leiping, spoken in southwestern Guangxi. Its tone system is given in Table 6-11 Table 6-11 Pattern of tonal splits and mergers in Leiping aspirated unaspirated glottalized voiced *A *B *C *DS *DL 1 = [454] 2 = [44] 3 = [13ˀ] 2 2 2 = [44] 5 = [11] 4 = [21] 45 The easiest tone to reconstruct for Leiping is *C because it does not show any split. Therefore we can posit an earlier glottalized rising tone for this Leiping tone. For tones *A and *B, it is apparent that A1 (= tones 1 and 2) and B1 (= tone 2) are 261 distinguished from A2 (= tone 4) and B2 (tone 5) respectively in terms of register. While tones A1 and B1 are high tones, tones A2 and B2 are low tones. Therefore, we can posit a falling tone and a level tone for Leiping *A and *B respectively. In addition, *DS and *DL are both connected to *B. We can consider tone 4 [21] in DS2 to have secondarily developed from the phonetically similar tone 5 [11] due to syllable structure. We can thus see that *DS in Leiping is also linked to *B. In sum, Leiping had a falling tone, a level tone, and a rising tone. The next variety to be discussed is the SWT dialect Siamese in Table 6-12. Table 6-12 Pattern of tonal splits and mergers in Siamese aspirated unaspirated glottalized voiced *A *B *C 5 = [24ˀ] 2 = [21] 3 = [42] *DS 2 *DL 2 1 = [33] 3 = [42] 4 = [453ˀ] 4 3 The mechanism of tone-segment interaction can also derive the modern Siamese five-tone system from the *ABCD system. In this case, the first-series tones consistently have lower onsets than their second-series counterparts. As tones B1 (= tone 2), B2C1 (tone 3), and C2 (tone 4) are all falling tones, we can posit falling tones for both *B and *C. Siamese *B thus must have been a mid falling tone, while *C must have been a high falling tone. In addition, *C must have been glottalized as indicated by the presence of glottalization in C2. Similarly, tone A1 (= tone 5) most likely resulted from lowering of the onset of tone A2 (= tone 1) due to the preceding aspiration, therefore indicating that Siamese *A was mid level tone. As for *D, we 262 again see a link between *DL and *B. *DS, however, does not seem to have had connection to *DL. However, if we consider tone 2 [21] in DS1 to have resulted from a reduction of tone 3 [42] due to syllable structure, we arrive at a new connection between *C and *DS. In sum, Siamese had mid level, mid falling, and high falling tones before the register split. The next variety to be discussed is Black Tai, a SWT dialect of northwestern Vietnam, given in Table 6-13. Table 6-13 Pattern of tonal splits and mergers in Black Tai aspirated unaspirated glottalized voiced *A *B *C *DS *DL 1 = [22] 2 = [45] 3 = [21ˀ] 2 2 4 = [55] 5 = [44] 6 = [31ˀ] 5 5 The Modern Black Tai tone system is a straightforward case of a binary register split. The second-series tones A2 (=tone 4), B2 (= tone 5), and C2 (=tone 6) can be viewed as a higher version of the first-series tones A1 (= tone 1), B1 (= tone 2), and C1 (= tone 3) respectively79. As tones A1 and A2 are all level tones, we can posit a level tone for the original Black Tai system. Similarly, both C1 and C2 are glottalized falling tones, indicating that Black Tai *C must have been a glottalized falling tone. In the case of *B, tone B2 can be considered tone B1 with a raised onset 79 Although voiced onsets usually induce lower pitch, cases of high tones from earlier voiced onsets are amply attested as discussed in Kingston and Solnit (1989) and Kingston (2004). 263 resulting from the register split80. Therefore, we can reconstruct Black Tai *B as a rising tone. Therefore, Black Tai *A, *B, and *C were a mid, a rising, and a falling tone, respectively. Notice that both *DL and *DS are tied to *B. The next variety to be discussed is the NT dialect of Wuming in central Guangxi given in Table 6-14. Table 6-14 Pattern of tonal splits and mergers in Wuming aspirated unaspirated glottalized voiced *A *B *C *DS *DL 1 = [33] 5 = [24] 3 = [55] 5 5 2 = [31] 6 = [13] 4 = [51] 6 6 The pattern of tonal splits and mergers in Wuming is identical to that of Black Tai but the contour shapes are very different. Columns A and C are parallel. Tones A1 (= tone 1) and C1 (= tone 3) are both level but C1 is higher. Similarly, tones A2 (= tone 2) and C2 (= tone 4) are both falling but C2 is higher. Therefore, we can reconstruct two falling tones for Wuming *A and *C respectively. The difference between the two must have been pitch height. *A must have been a mid tone while *C must have been a high tone. As for *B, both the first-register and second-register reflexes, i.e. B1 (= tone 5) and B2 (= tone 6) respectively, are rising tones in the lower pitch region. Therefore, we can reconstruct Wuming *B as a low rising tone. From a 80 In Gedney’s impressionistic transcription, tone B1 is described as high rising [45]. To consider tone B2 as tone B1 with a raised onset, we need to speculate that this high rising tone in Black Tai may in fact be something between [34] and [45]. 264 systemic point of view, the original contrastive property of Wuming tones seem to have been pitch height—mid for *A, low for *B, and high for *C. In addition, the identity between *B and *D in modern Wuming suggests that the two tones were the same tone, or had properties in common. Now consider the NT dialect of Liujiang in the northeast of Guangxi, given in Table 6-15. Table 6-15 Pattern of tonal splits and mergers in Liujiang aspirated unaspirated glottalized voiced *A 1 = [42] *B *C *DS 5 = [33] 3 = [53] 381 *DL 5 2 = [231] 6 = [22] 4 = [24] 4 6 Liujiang’s tonal system is quite similar to that of Wuming both in terms of splits and mergers and in terms of the actual tone shapes. Like Wuming, A1 (= ton3 1), and A2 (= tone 2) are both falling tones, even though A2 is in fact rising-falling. We can thus posit a falling tone for Wuming *A. However, Liujiang differs from Wuming in that C1 is a falling tone but C2 is rising. Therefore, it is not clear what tone to posit for *C. As for *B, Liujiang has level tones both for B1 and B2, indicating an earlier level tone. Again, *B is linked to *DL as in most modern Tai dialects. Moreover, there is a very clear connection between *C and *DS, as in Siamese. The 81 Zhang et al. (1999) describe Liujiang DS1 as [55] and follow the Chinese tradition in considering this checked-syllable tone as a separate tone. However, I view [55] as a checked-syllable allophone of tone 3 [53]. 265 last tone system to be examined is that of Lianshan, a NT variety spoken in Guangdong, provided in Table 6-16. Table 6-16 Pattern of tonal splits and mergers in Lianshan aspirated unaspirated glottalized voiced *A *B *C 1 = [42] 5 = [35] 3 = [55] *DS *DL 35 2 = [132] 6 = [13] 4 = [214] 6 6 The Lianshan tone system shows both the basic binary register split and the quantity-based split in *D syllables. Both tone A1 (= tone 1) and A2 (= tone 2) are falling tones but they differ in terms of onset. While the former starts in the upper part of the pitch range, the latter starts very low and rises before falling slightly. This early low pitch possibly resulted from lowering through the mechanism of segment-tone interaction. Therefore, we can posit a falling tone for *A. The reconstruction of Lianshan *B is straightforward since both B1 (= tone 5) and B2 (= tone 6) are rising tones. Therefore, the *B tone must have been a rising tone in an earlier stage of Lianshan. As for tone *C, it is not clear what tone shape it may have had because C1 (= tone 3) and C2 (= tone 4) agree neither in pitch height or contour shapes. Like many other Tai dialects, the there is a connection between *B and *DL. In addition, the link between *C and *DS identified for Siamese and Liujiang is also found in Lianshan. Although we expect tone 4 for DS2 rather than tone 6, but it is very likely that tone 6 [13] in short checked syllables resulted from a contour reduction of tone 4 [214]. 266 In this section, I have applied internal reconstruction to individual dialects to arrive at the tone system before the binary register splits occurred. We will next use the Comparative Method to arrive at the formal characteristics of the PT tonal categories. In the following section, the seven *ABCD systems will be compared in order to infer the characteristics of the common ancestral *ABCD system. 6.5 Reconstructing the contrasts Having internally reconstructed the shapes of the *A, *B, *C, and *D tones in individual varieties, the next step is to compare these reflexes to arrive at a concrete reconstructions the PT tonal categories. In order to put forward a proposal about the nature of the PT tonal contrast, I first reconstruct the following aspects of the PT tonal categories: pitch height, pitch contour, vowel duration, and voice quality. Then, I proceed to discuss the four phonetic parameters and the claim that PT tones should in fact be considered register complexes. 6.5.1 Comparing *ABCD in daughter languages In this sub-section, I propose a concrete reconstruction of PT tones by comparing reflexes of PT tones in individual varieties before the binary split as established by internal reconstruction in §6.4. In comparing the reflexes, if all the varieties agree, the choice of reconstruction is obvious. In contrast, if the languages present two possible reconstructions, other considerations come into play. For example, by comparing modern reflexes alone, *A can be either falling or level but I posit a level contour for *A because *C is clearly falling (see below). Lastly, if the languages present more than two possibilities, no conclusion can be made. The first PT tone to be considered is *A, whose reflexes in individual varieties discussed earlier are given in Table 6-17. 267 Table 6-17 PT *A in individual varieties Siamese height contour voice quality mid level Black Tai level Cao Bang high falling Leiping falling Wuming mid Liujiang Lianshan falling falling The reflexes of PT *A in individual Tai varieties suggest two characteristics of this proto-tone. First, PT *A must have been either a mid or a high tone. In other words, it must have been in the upper region of the pitch range. This is because varieties whose pre-register split reflexes can be internally reconstructed either had a mid or a high tone for *A, never a low tone. Second, the contour of PT *A must have been either level or falling as reflected in individual varieties except for Wuming. None of the varieties discussed show a rising tone. The situation for tone *B is very different as summarized in Table 6-18. Table 6-18 PT *B in individual varieties Siamese height contour voice quality mid falling Black Tai rising Cao Bang mid falling Leiping level Wuming low Liujiang Lianshan level rising 268 The reflexes of *B in individual varieties only point to possible pitch heights but not contours of PT *B. Individual dialects had either mid or low tones as reflexes for PT *B. In other words, they indicate that PT *B occupied the lower range of the pitch range. As for tonal contour, there is no clear tendency among individual dialects. Level, falling, and rising contours are attested among individual varieties. Therefore, the only characteristic of PT *B that we can posit at this point on the basis of reflexes in the daughter dialects compared is the pitch height. The case of *C, summarized in Table 6-19, is clearer. Table 6-19 PT *C in individual varieties Siamese height contour voice quality high falling ˀ Black Tai falling ˀ Cao Bang mid rising Leiping Wuming Liujiang Lianshan rising ˀ high Like PT *A, tone *C was reflected either as a high or a mid tone in the preregister split stage of individual languages. This suggests that tone *C must have occupied the upper range of the pitch range. As for the contour shape, individual varieties show falling or rising reflexes, suggesting that the PT tone *C may have a dynamic tone, not a level tone. In addition, Gedney (1989d)’s observation that reflexes of *C tones are often glottalized is true for the pre-register split stage of the dialects 269 discussed. This fact supports his proposal that PT *C had glottalization 82 as an important characteristic. Although we cannot make completely conclusive characterizations of the shapes of PT tones at this point, non-trivial relationships among the categories have become apparent. First of all, tones *A and *C must have been higher in pitch than tone *B because they most likely occupy the upper part of the pitch range. The *B tone, in contrast, must have been relatively low in pitch. Second, we also see a difference in terms of pitch contour among the PT tones. *A is unlikely to go back to a rising tone while *C could not have been a level tone. Moreover, we also see that, at least for one tone, voice quality was an important property. Specifically, tone *C mostly was characterized by glottalization. Based on the result of the comparison above, I now put forward a reconstruction of the phonetic characteristics of the PT tonal categories. The three contrastive tones in the PT *ABCD system differed from each other in four dimensions: 1) pitch level, 2) tonal contour, 3) vowel duration, and 4) voice quality. A proposed reconstruction of PT *A, *B, *C, and *D is presented in Table 6-20 . 82 Gedney (1989d) seems to view this glottalization as a segment *-ʔ while I view it as part of the tone. 270 Table 6-20 Phonetic characteristics of PT tones pitch height contour vowel duration voice quality *A mid level modal *B low rising long creaky *C high falling short glottal constriction *D low rising I propose that PT *A was a modal tone with mid level contour. In contrast,*B was a creaky tone with a low rising contour, and relatively long vowel duration. Lastly, *C had a high falling contour ending in a glottal constriction, and a relatively short vowel duration. Again, *D is viewed as belonging to the same tone as *B. In the following sub-sections, I discuss each of the phonetic properties that characterized the PT tonal categories. 6.5.2 Vowel duration In tonal system, different tones co-vary with duration of the vowels (Abramson 1962; Gandour 1977; Kong 1987; Yu to appear). For PT tonal categories, the tonogenetic behavior of *D reveals more relationships of just such a systemic nature. I propose that PT *B had a relatively long vowel duration but PT *C was relatively short. Table 6-21 summarizes the connection of *D to *B and *C in the individual varieties. 271 Table 6-21 Connection of *D to *B and *C Siamese Black Tai Cao Bang Leiping Wuming Liujiang Lianshan *DL *B *B *B *B *B *B *B *DS *C *C *C As seen above, PT *D has connections with both *B and *C. The strongest link is that between *B and *D. In general, *D followed the same pattern of splits and mergers as *B. The clearest case are Black Tai, Leiping, and Wuming, where *D belongs to the same tone as *B synchronically. In cases where *D split into *DL and *DS, *DL patterns with *B as is the case for Siamese, Cao Bang, Liujiang, and Lianshan. As Gedney discusses (1989d), this suggests that tone *D was phonetically very similar to the *B tone at the PT level (see §6.2). As for *DS, it patterns with *C in many cases suggesting that when *D split into two tonal categories, *DS came to be identified with *C instead of *B. These inter-relationships among *B, *C, and *D reveal that duration may have played an important role in the contrast between *B and *C. If *B and *DL were phonetically similar, and *C and *DS were similar, then the difference between *B and *C on one hand and the difference between *DL and *DS on the other must be of the same nature. Since the difference between *DL and *DS is clearly that of vowel length, the difference between *B and *C therefore may have been one of duration. In other words, PT *B must have been in some way longer than *C. 272 Because PT is proposed to have had a vowel-length contrast in closed syllables, vowel duration, which functioned as one of the phonetic correlates of tones, must have interacted directly with contrastive segmental vowel length. In non-checked syllables *A, *B, and *C, the duration of the phonologically short vowels was probably roughly half the duration of the phonologically long vowels. When comparing across tones, the short and the long vowels with tone *B were longer than their respective counterparts with tone *C. Most crucially, long vowels that were shortened due to *C were still significantly longer than short vowels in any of the tones. Similarly, lengthened short vowel with tone *B were significantly shorter than long vowel with any of the tones. This type of interaction between tone and vowel length is exactly what we find in Siamese (Abramson 1962) illustrated in Table 6-22. Table 6-22 Siamese tones correlated with average vowel durations in msec (Abramson 1962: 107)83 low tone high tone short vowel 163 97 long vowel 524 436 The Siamese data in Table 6-22 presents a comparison between Siamese long and short vowels in syllables with the low tone (= B1) and the high tone (= C2). It shows clearly that the phonetic vowel duration that functions as a property of tones does not interfere with the phonological vowel length contrast. Short vowels and long vowels are kept distinct from each other on syllables with different tones. For any given tone, the duration of long vowels in Siamese is more than twice as great as the duration of their respective short counterparts. 83 Only the data for speaker W.N. are shown here. 273 The situation may have been a little different for checked syllables, or tone *D. First of all, bear in mind that PT *D did not contrast for tones. Second, it is not uncommon for vowels in checked syllables to be phonetically shorter than vowels in non-checked syllables (Gordon 2001; Zhang 2002). Therefore, it is not surprising that a phonologically short vowel in a *D syllable would have roughly the same duration as the short vowel in a *C syllable. It is also not impossible for the duration of a long vowel on *D syllables to be so short as to resemble that of a short vowel on *B syllables. This kind of situation would give rise to a length-based tonal split in which *DS patterns with *C and *DL with *B, as is observed among Tai dialects. Such split is exactly what we happened in so many Tai varieties including Siamese, White Tai, Western Nung etc. (see §6.2). 6.5.3 Pitch height Pitch heigh is one of the most important phonetic properties of tones. The pitch height characteristics of the PT tonal categories have been partially uncovered by comparing their reflexes in daughter dialects (see §6.5.1). The comparison reveals that *C had a higher pitch than *B. In addition, the pattern of tonal splits and mergers also provides crucial information. In particular, the behavior of *D with respect to tonal mergers and splits points to a shorter duration for *C than for *B (§6.5.2). Based on this intriguing pitch-duration patterning, I propose that the pitch heights of PT *A, *B, *C were in the mid, low, and high respectively. As for *D, I claim that pitch height was not among its distinctive phonetic properties. This patterning of higher pitch with short vowels on one hand, and lower pitch with long vowels on the other, is a very common phenomenon cross-linguistically. Diachronically, this pitch-duration patterning has been shown to be the source of tonal contrasts in two Mon-Khmer languages: U (Diffloth 1991) and Hu (Diffloth 1991; 274 Svantesson 1991). In these cases, low tones have been shown to have developed from long vowels while high tones developed from short vowels. Synchronically, various studies on Thai (Abramson 1962), English (Lehiste 1970), Kui (L-Thongkum 1989), and Chong (L-Thongkum 1991) among others have shown that vowel duration inversely relates to the pitch. Therefore, positing a low and high tone respectively for PT *B and *C nicely accounts for both the fact that *B had a lower tone than *C, and the fact that the *B was somehow longer than *C. As for PT *A, I posit a mid tone for two reasons. The first reason is that dialect data suggest that *A was higher than *B. Therefore, it must have been either a mid or a high tone. However, treating *A as a mid tone is more likely because *C is already a high tone. As for the vowel duration, *A does not usually pattern with either *DS or *DL with respect to tonal splits and mergers. Therefore, it is unclear whether the height of *A had an effect on the duration of the vowel. As for *D, the fact that it is a checked-syllable variant of *B suggests that it also occupied the lower part of the pitch range. 6.5.4 Pitch contour In addition to pitch height and vowel duration, the comparative data discussed in §6.5.1 also suggest that pitch contours were important phonetic properties of PT tones. Here I propose that PT *A, *B, *C, and *D had level, rising, falling, and rising contours respectively. If we base our analysis on the contours in individual daughter dialects alone, PT *C could be reconstructed with either a rising or a falling contour. Similarly, there is no agreement among dialects whether tone *A had a level, or a falling contour. However, the glottalization, which is clearly indicated by the comparative data, favors a falling contour for *C. Phonetically, a glottal closure may have an effect of either 275 raising or lowering of the pitch of the preceding vowel (e.g. Hombert, Ohala, & Ewan 1979; Mazaudon 1977; Thurgood 2002b). This means that a level contour is not a likely candidate for a glottalized tone. Therefore, I posit a falling contour for *C, reserving level contour for *A. Since *A and *C are reconstructed with a level contour and a falling contour respectively, we can thus posit a rising contour for PT *B. This hypothesis for *B is supported by two facts. First, languages that point to a rising contour in *B are spoken very far apart. Black Tai is a SWT dialect spoken in the northwest of Vietnam while Lianshan is a NT dialect spoken in the northwest corner of Guangdong. This makes the rising contour seem like a case of retention. Second, both experimental and typological studies (Gordon 2001; Kong 1987; Yu to appear; Zhang 2002) have shown that vowels under rising tones tend to be longer than vowels on level and falling tones. The longer duration of tone *B discussed earlier is thus very suggestive of a rising contour. Therefore, I posit a rising contour for PT *B. Finally, I also reconstruct a rising contours for *D because of its identity to *B. One may argue that, typologically, checked syllables with rising contour are typologically marked (Gordon 2001; Zhang 2002), casting doubt on the view that *D had a rising contour. However, many modern Tai dialects attest rising tones on checked syllable, cf. Wuming, Lianshan, and Cao Bang in §6.4. Wuming, crucially, has only rising tones on checked syllables. Moreover, the reflexes of *DS in numerous modern dialects, e.g. Tai Yuan, the Lao dialect of Nong Khai, Yay, Liujiang, etc., are rising tones. Given the typological restriction against rising tones on checked syllables, it is very implausible that these diverse daughter languages would independently developed rising contours on checked syllables with short vowels. Therefore, the rising contours in reflexes of tone *D must be considered a retention from PT. 276 6.5.5 Voice quality The last phonetic property of the PT tones to be discussed is voice quality. In addition to the glottalization of tone *C that is robustly supported by the reflexes in modern dialects, we can hypothesize about the voice quality of PT *A and *B. I propose that PT *A, *B, *C, and *D had modal, creaky, finally glottalized, and creaky voice qualities respectively. Before proceeding to discuss the voice qualities of PT tones, a distinction between “creakiness” and “final glottal constriction” must be made. “Creakiness”, or “laryngealization”, refers to the irregular, low-frequency vibration of the vocal folds extended over a relatively long period of the duration of the vowel (Gordon & Ladefoged 2001; Michaud 2004). It is produced with the vocal folds tightly adducted but open enough to allow voicing. In contrast, “final glottal constriction” refers to a complete or almost complete closure of vocal fold at the end of the rime (Michaud 2004). The vocal fold vibration of the vowel is not irregular. The differences between the two is demonstrated by Michaud (Michaud 2004: 137-139), who clearly compares the “broken” and the “drop” tones in Vietnamese. While the former is creaky, the latter is characterized by a glottal constriction. The reconstruction of voice quality for *A is straightforward because there is no evidence to date that would indicate a special voice quality. Therefore, it is likely that at the PT level *A had a modal voice quality. For *C, reflexes in modern dialects robustly indicate that it must be reconstructed with a characteristic glottalization (see §6.5.1). However, the nature of the glottalization needs clarification. In most dialect descriptions, the term glottalization is never explicitly defined. However, it is clear from Gedney (1989d: 210)’s discussion that the glottalization found in reflexes of *C is not creakiness but a final glottal constriction preceded by a transitional period of 277 intermittent voicing following modal voicing. Therefore, I propose that PT *C characteristically had a final glottal constriction. The regular correspondence of *C to the Rising tone in Chinese also lends additional support to the reconstruction of final glottal constriction as part of PT *C. This is because the Rising tone is believed to have developed from OC final glottal stop (Baxter 1992; Haudricourt 1954b) and to have modal voice quality ending in a glottal stop in MC (Sagart 1986). Additional support for creakiness in *B comes from the durational difference between PT *B and *C. That PT *B had a relatively long duration while *C had a relative short duration suggests that they had creaky voice and final glottal constriction respectively. As discussed by Michaud (2004), final glottal constriction tends to end the syllable abruptly, resulting in a shorter vowel duration. For example, Vietnamese rimes carrying the ‘drop’ tone are impressionistically described as “immediately cut off by a strong glottal stop (Thompson 1965: 41). Acoustically, the “drop” tone is shorter than the other tones and ends in a glottal stop (Pham 2003). Similarly, Burmese tone 4, which developed from earlier /-ʔ/, is extra-short phonetically (Bradley 1982). Therefore, the association of *C with final glottal constriction fits together very nicely with the shorter duration of the proto-tone. As for *B, the reconstruction of voice quality is not unequivocal. That *C is reconstructed with final glottal constriction and *A with modal voice suggests that *B might have also had a characteristic voice quality at the PT level. Therefore, I hypothesize that PT *B had a creaky voice quality. Support for this hypothesis comes from a few dialects not included in §6.4 and §6.5.1. Dialects that show non-modal voice quality for reflexes of PT *B are found in three geographical areas: the Shanspeaking area, Lao-speaking area, and the CT area of northeastern Vietnam. Various dialects from these areas have been described as having glottalized reflexes of PT *B. In the Shan-speaking area, we have the dialects of Hsi Paw (Strecker 1979), Kengtung 278 (Hudak 1994), and Hsenwi (Strecker 1979) in Myanmar; the dialect of Mae Sot (Sarawit 1979) in Thailand; and the dialect of Chefang in Yunnan (Hudak 1994). In the Lao-speaking area, we have the Nyo dialect of Tha Uthen in northeast Thailand (Hudak 1997). Lastly, in northeast Vietnam, we have Nung Fan Slihng (Freiberger 1964; Freiberger & Bé 1976; Saul & Freiberger Wilson 1980) and two other CT dialects from Po Muc (Hudak 1995), and Bac Va (Hudak 1995)84. Because the dialect descriptions usually do not distinguish between the two types of glottalization, i.e. final glottal constriction and creaky voice, we can only conclude that these dialects show glottalized voice quality in tones that developed from PT *B. Another complication is that the creak-like quality in B tones in some dialects such as tone B1 in Chefang may in fact have developed secondarily as an enhancement property of low tones, cf. Mandarin tone 3 (Belotel-Grenié & Grenié 1994; Davison 1991). Among the varieties mentioned, Nung Fan Slihng, shown in Table 6-23, is the strongest piece of evidence for the creakiness of tone *B due to the clarity of its description (Freiberger 1964; Freiberger & Bé 1976; Saul & Freiberger Wilson 1980). 84 Sagart (1988) was the first to use the glottalized tones in Hsi Paw Shan, Mae Sot Shan, and Nung Fan Slihng as evidence for reconstruction of PT tones. Also see §6.2 for arguments against his proposal that *B had a final glottal stop. 279 Table 6-23 Pattern of tonal splits and mergers in Nung Fan Slihng *A aspirated 1 low-rising unaspirated, vcl. 4 sonorants high-rising glottalized voiced 3 mid level *B 6 4 6 low glottal *C 5 low falling *DS 4 1 2 high glottal 3 *DL 6 4 6 As shown above, the reflexes of PT *B2 and *C2 in Nung Fan Slihng are both glottal tones. As expected, the C2 tone ends with glottal constriction. It is described as having a “high rising glide and a final glottal stop (Freiberger 1964).” Tone B2 ends with a glottal constriction similar to C2, but additionally shows creakiness in the vowel. According to the description, it “ends with a final glottal stop, but laryngealization may occur over all the voiced phonemes (Freiberger 1964).” I claim that this optional laryngealization in tone B2 is a remnant of the creaky voice characteristic of PT *B. It is also possible that the laryngealization is a synchronic enhancement property of low tones which does not have anything to do with PT voice quality. However, the fact that the low falling tone, i.e. C1, is not also described as showing laryngealization suggests that the creakiness is not an enhancement feature of low tones. Further support for the creakiness of PT *B comes from the regular correspondence to Chinese Departing tone (see §6.2). The most obvious explanation for the correspondence between PT *B and the Chinese Departing tone is that were phonetically similar. There is a consensus among Chinese historical linguists that the Departing tone in MC developed from OC final *-h from an earlier *-s (Baxter 1992; 280 Haudricourt 1954b; Pulleyblank 1991; Sagart 1999). However, opinion varies with respect to the intermediate stage between non-tonal OC and the modern Chinese dialects which are fully tonal. Haudricourt (1954a) and Pulleyblank (1984) believe that the intermediate stage of the Departing tone was a final aspiration. In contrast, Sagart (1986; 1988) argues that the Departing tone in MC was a short tone with creaky phonation. Reconciling these hypotheses about the development of the Departing tone, we arrive at the following evolutionary path: *-h (from earlier *-s) > creaky or aspirated > pitch-based tone. Together with the glottalized reflexes of *B in modern dialects, the correspondence between PT *B and the Departing tone can therefore be taken as additional support for creaky voice in PT *B. Additional support for creakiness in *B comes from the durational difference between PT *B and *C. That PT *B had a relatively long duration while *C had a relative short duration suggests that they had creaky voice and final glottal constriction respectively. According to Gordon and Ladefoged (Gordon & Ladefoged 2001), nonmodal voice qualities tend to increase the duration of vowels cross-linguistically. For example, Jalapa Mazatec (Silverman et al. 1995) creaky vowels are longer than their modal counterparts. Similarly, the rime of Vietnamese “broken” tone syllable is creaky and has greater duration. Therefore, the association of *B with greater vowel duration supports the hypothesis that *B was a creaky tone. Further, I propose that PT *D did not have a voice quality characteristic of the tone itself. Instead, I claim that voice quality was neutralized in checked syllables. However, I believe that the vowel in *D syllables were phonetically glottalized due to the following final stops. This glottalization explains why *B and *D patterned together. In particular, glottalization in *D resembled the creakiness characteristics of *B. As discussed by Hombert et al. (1979), Mazaudon (1977), and Thurgood (2002b), 281 glottalization of vowels preceding final stops has been reported for numerous languages of SEA, e.g. Khaling, Kiranti etc. 6.5.6 PT as a “pitch-plus voice quality” tone language I have proposed above a concrete reconstruction of PT tones *A, *B, *C, and *D. In the proposal, PT tone *A had mid level pitch contour with modal voice. In contrast, tone *B had a low rising contour with by a relatively long vowel duration, and creaky voice. *C had a high falling contour with short vowel duration, and final glottal constriction. Lastly, *D, which contrasted with the other tones in terms of syllable structure, had a low rising pitch contour similar to *B. Not coincidentally, this proposed PT tone system resembles closely that proposed for Proto-Kra (Ostapirat 2000). This bundling of phonetic features closely resembles the situation in Northern Vietnamese. In this dialect of Vietnamese, tones are not distinguished solely by pitch but are complex bundles of pitch contour and voice quality (Brunelle 2009; Maspéro 1912; Michaud 2004; Pham 2003). For example, the “broken” tone shows salient laryngealization, while the “drop” tones are characterized by a final glottal constriction. Moreover, the “broken” tone has a greater rime length in contrast with a relatively short rime length in the “drop” tone. Therefore, I claim that PT, like Vietnamese, was a “pitch-plus-voice quality” tone language85. 6.6 From PT tones to modern reflexes Having proposed a concrete reconstruction of PT tones, I now illustrate how modern reflexes can be derived from the reconstructed tones. Based on the framework outlined in §6.3, the discussion will be limited to the seven varieties analyzed in §6.4: Cao Bang, Leiping, Black Tai, Siamese, Wuming, Liujiang, and Lianshan. 85 I adopt this term from Michaud (2004). 282 As discussed earlier, the *A tone was a modal tone with a mid level contour in PT but is reflected as a level or falling tone in daughter languages. In Siamese and Black Tai, PT *A was kept as a level tone before the register split. In contrast, Cao Bang changed tone *A to a high falling tone before the register split. As for Liujiang, Lianshan, and Leiping, *A tone was also a falling tone but the pitch height is unknown. Perhaps, this is because the level contour of PT *A was in fact slightly falling as is the case for the modern Siamese mid tone (Morén & Zsiga 2006). This phonetic fall may have been reanalyzed in these four varieties so that *A became a mid falling tone in these four dialects. The high falling contour is Cao Bang can then be accounted for by the mechanism of perceptual maximization (§6.3.3), which raised the onset from mid to high so that the fall is perceptually more distinct. PT *B is reconstructed here as a creaky tone with a low rising contour. However, the creakiness has been completely lost in all the varieties discussed here. Moreover, its reflexes in individual varieties vary tremendously. Black Tai and Lianshan both have a rising tone, which can be regarded as a case of retention. Leiping and Liujiang had a level tone, which may have been derived from the PT rising contour by contour reduction (§6.3.2.1). That is, the rising contour was flattened, yielding a level contour. For Siamese and Cao Bang, the *B tone had a falling contour before the register split, the opposite of the original PT rising contour. We can account for the radical change by two simple steps, keeping in mind that a tone with a low onset usually starts in the mid region before falling to reach the low target (Xu 2004). The first step was the loss of pitch rise through peak sliding (§6.3.2.2). This mechanism left *B with a simple fall that started in the mid region. In the next step, the fall may have become reanalyzed as a distinctive characteristic of *B. As for *C, I reconstruct a high falling contour with a final glottal constriction. Wuming kept this PT tone as a high tone but it is not clear what the contour was like. 283 In modern Wuming, the glottal constriction has been lost. The CT dialects of Leiping and Cao Bang both had rising tones for *C. The change can be explained in a few steps. First, the initial rise, which was originally, a transition from the mid region, became more prominent as the peak was realized later in the rime through the mechanism of peak delay. After this prominent rise became phonologized, it became a target of perceptual maximization such that the tone onset was lowered to create a steeper pitch rise, yielding a true rising tone. This series of changes is attested in the case of modern Siamese tone 4 (Pittayaporn 2007b; Theranon 2002a, 2002b). Note that Cao Bang has lost its original glottal constriction while Leiping maintains it. As for Siamese and Black Tai, both the falling contour and glottal constriction were still present at time of the register split. PT *D was phonetically similar to *B even though the truly contrastive characteristic was the syllable structure. This means that *D had a low rising contour as discussed above. However, in most dialects *D later split into *DS and *DL based on vowel length. *DS, which had short vowels at the time of this split, remained unified with *DL and *B in some dialects including Black Tai, Leiping, and Wuming. In contrast, it was re-associated with *C in many dialects including Siamese, Liujiang, and Lianshan. In these varieties, *DS came to be the same tone as *C because its short vowel resembled the relatively short vowel duration characteristic of *C. In addition, there are also some cases where *DS possibly became associated with *A, e.g. Cao Bang, Leiping, and Southern Shan86. In contrast to *DS, *DL, which developed from checked syllables with long vowels, in most cases remained the same tone as *B. The seven varieties used as representatives in §6.4, including Siamese, Black Tai, Cao Bang, Leiping, Wuming, 86 In Yay and most NT varieties, the shortening of high vowels before stops (see §5.6.1) occurred before the length-based split occurred, as indicated by the fact that etyma that had these rimes in PT all show DS tones rather than DL tones in these languages. 284 Liujiang, and Lianshan, all retained the connection until the binary register split. In a handful of languages, however, there is no clear trace of such a link. These varieties include Lao dialects in Laos and northeastern Thailand, as well as Southern Tai dialects in Southern Thailand and northern Malaysia. In this section, I have shown that the proposed concrete reconstruction of PT tones can straightforwardly account for the wide range of modern reflexes using a framework of tone change that was independently motivated. This is a good indication that the propsoed reconstruction is a plausible one. 6.7 Summary Although it has long been established that PT had four tonal categories *A, *B, *C, and *D, this chapter constitutes a major advancement in uncovering the specific phonetic content of the four PT tonal categories. In particular, I proposed that PT was a tone language that made use of both pitch and voice quality for tonal contrast. I posited a modal tone with a mid level contour for *A; a creaky tone with a low rising contour and a relatively long vowel duration for *B; and a high falling contour ending with a glottal constriction with a relative short vowel duration for *C. Lastly, *D patterned with *B because they were phonetically similar. Specifically, I posited a low rising contour for *D. Not only is the proposed tonal system very plausible from a phonetic and areal point of view, it can also account for the range of variation among reflexes in different modern dialects in a simple way. 285 CHAPTER 7 DISCUSSION 7.1 Introduction The reconstruction of PT that I have motivated in this dissertation holds that PT was a sesquisyllabic language in which phonological words were either monosyllables or sesquisyllables. Diachronically speaking, this stage is one step away from the monosyllabicism observed in modern Tai languages. This view of PT phonology differs markedly from Li’s reconstruction (1997) and offers a very different picture of the historical development of Tai languages. This chapter discusses the implications of this reconstruction. I first compare the proposed reconstruction of PT phonology with Li’s reconstruction. Subsequently, I propose a preliminary subgrouping schema of Tai languages. I then relate my reconstruction of PT phonology to the most recent findings about Kra-Dai historical phonology, and also discuss two factors that contribute to the convergence among modern Tai dialects. 7.2 Comparison of PT reconstructions As discussed in Chapter 1, Li’s reconstruction of PT (Li 1977) holds that PT was a monosyllabic language with a rich set of onsets, and a massive vowel system, but a relatively impoverished set of codas. As for tone, previous reconstructions only provide abstract labels for the reconstructed tonal categories. The current reconstruction differs markedly from that conventional view of PT phonology as it claims that PT was a sesquisyllabic language that had both monosyllabic and sesquisyllabic prosodic words. This view leads to a very different reconstruction of PT onsets and rimes, as we have seen. Moreover, the current system proposes a concrete reconstruction of the tonal categories established by previous research. In this section, 286 I provide a comparison between the current reconstruction and the heretofore prevalent view of PT phonology. According to previous reconstructions, PT had a rich consonant system consisting of a four-way phonation type contrast among stops and five contrastive places of articulation. The current reconstruction, on the other hand, posits only three contrastive phonation types, but posits six contrastive places of articulation. With respect to phonation types, the current system reconstructs plain voiceless, implosive, and voiced stops, but lacks the aspirated stop series that is central to previous reconstructions. For example, Li (1977) reconstructs altogether four labial stops *p, *pʰ, *ˀb, and *b, but the view of PT phonology adopted here only posits three: *p, *ɓ, and *b. As for places of articulation, my reconstruction posits a distinctive uvular series, in addition to the labial, alveolar, palatal, velar, and glottal series proposed in earlier reconstructions. For example, while Li reconstructs four different plain stops *p, *t, *k, and *ʔ, plus an affricate *tɕ, the current reconstruction of PT phonology posits altogether six voiceless stops *p, *t, *c, *k, *q, and *ʔ. As for manner of articulation, the current and previous reconstructions agree in having stops, fricatives, nasals, liquids, and glides, but differ in three respects. First, the current system treats the palatal affricates of previous reconstructions as stops: *c and *ɟ. Second, it lacks the labial fricatives *v, and *f but has a uvular fricative *χ and possibly palatal fricatives *ɕ and *ʑ. Fthe palatal glide *j does not exist in the current system. The only palatal glide in the system is *ˀj but it belonged to the glottalized series rather than the sonorants. Note that *w- and *ʰw- in the current proposal do not correspond to *w- and *ʰw- in previous reconstructions. Table 7-1 gives the PT consonants posited in the current reconstruction and their equivalents in Li’s reconstruction (1977) as well as alternative proposals (Ferlus 1990; Sarawit 1973). 287 Table 7-1 PT consonants and their correspondents in previous reconstructions PT 1 *p2 *t3 *c4 *k5 *q6 *ʔ7 *ɓ8 *ɗ9 *ˀj10 *b11 *d12 *ɟ13 *g14 *ɢ15 *s16 (*ɕ-) 17 *x18 *χ19 *h20 *z21 (*ʑ-) 22 *ɣ23 *ʰm24 *ʰn25 *ʰɲ- Li Sarawit Ferlus *p- *p- *p- *t- *t- *t- *tɕ- *c- *c- *k- *k-, *kj*kʰ- *kʰ- *k*ŋk2- *ʔ- *ʔ- *ʔ- *ˀb- *ˀb- *ɓ- *ˀd- *ˀd- *ɗ- *ˀj- *ˀj- *ˀj- *b- *b- *b- *d- *d- *d- *ɟ- *ɟ-, *ɟw- *ɟ- *g- *g- *g- *ɣ- *ɣ- *g-/*ɣ- *s- *s- *s- *tɕ- *c- *c- *x- *x*x- *x- *x*ŋx2- *h- *h- *h- *z- *z-, *zw- *z- *dʑ- *ɟ- *ɣ- *ɣ- *ɣ- *ʰm- *ʰm- *ʰm-, *C.m- *ʰn- *ʰn- *ʰn- *ʰɲ- *ʰɲ- *ʰɲ- 288 Table 7-1 (Continued) PT Li Sarawit 26 (*ʰŋ-) *ʰŋ- *ʰŋ- 27 *m- *m- *m- 28 *n- *n- *n- 29 *ɲ- *ɲ- *ɲ-, *ŋj- 30 *ŋ- *ŋ- *ŋ- 31 *ʰw- *f- *f- 32 *ʰr- *ʰr- *ʰr- 33 *ʰl- *ʰl- *ʰl- 34 *w- *v- *v- 35 *r- *r- *r- 36 *l- *l- *l-, *lw- Ferlus *ʰŋ*m*n*ɲ- *f*ʰr*ʰl*v*r*l- As in Li’s reconstruction, two or more consonants can combine to form complex onsets. The biggest difference with regard to complex onsets is that the current system, like Ferlus’s, posits sesquisyllabic clusters while Li’s reconstruction allows only tautosyllabic clusters. This means that many sesquisyllabic clusters in this proposal correspond to tautosyllabic clusters in Li’s reconstruction. For example, *p.tin the current system corresponds to *tr- in Li’s reconstruction. The PT tautosyllabic clusters that I have posited are given in Table 7-2 together with their equivalents in Li’s (1977) and other reconstructions (Ferlus 1990; Sarawit 1973). 289 Table 7-2 PT consonants and their correspondents in previous reconstructions PT 1 *pr2 *tr3 *cr4 *kr5 *qr6 *br7 *gr8 *pl9 *kl10 *ɓl11 *gl12 *pw13 *bw14 *mw15 *ʰmw16 *kw17 *xw18 *ɣw19 *ŋw20 *qw21 *ɢw22 *sw23 *rw24 *tw25 *nw- Li *pʰl/r*tʰr*tʰl*xr*kʰl*br*gr*pl*kl*ˀbl/r*gl*pɰ- *b*mw*ʰmw*kw*xw*ɣw*ŋw*xw*ɣw- *s*r-, rw- *t- Sarawit *pʰl/r*tʰr*tʰl*xʰr*kʰl*br*gr*pl*kl*ˀbl*gl- *p*b*mw*ʰmw*kw*xw*ɣw*ŋw*xw*ɣʰw*swj*rw-, *rwj- *nw- Ferlus *pʰr*tʰr*cʰr*xr*kʰr2*br*gr*pl*kl*ɓl*ɣl*p*b- *kw*xw*ɣw*ŋw*xw*ɣw- 290 Table 7-2 (Continued) PT 26 *ʰrw27 *ʰlw28 *krw29 *klw30 *qrw31 *trw32 *crw- Li *ʰr*ʰl*kʰrw*kl*xr*tʰr*tʰl- Sarawit *ʰrw-, *ʰŋrw- *ʰlw*kʰlw*klw*xʰrw*tʰrw*tʰlw- Ferlus *kʰrw- PT sesquisyllabic onsets posited in the current reconstruction and their equivalents in Li’s reconstruction (1977) as well as alternative proposals (Ferlus 1990; Sarawit 1973) are provided in Table 7-3. Note that it is not possible, at least at this stage, to give an exhaustive list of sequisyllabic clusters in PT. Table 7-3 Examples of PT sesquisyllabic clusters and their correspondents in previous reconstructions PT Li Sarawit Ferlus 1 *p.t- *tr- *tr- *p.t2- 2 *k.t- *tl- *tl- *k.t- 3 *p.q- *x- *xw- 4 *q.p- *x- *x- 5 *q.t- *tr- *tr- 6 *C̥ .b- *b- *ˀbw- *ɓ- (<*ˀv-) 7 *C̥ .d- *ˀdl/r- *ˀdl- *ɗ- (<*ˀr-) 8 *C̥ .ɟ- *ˀj- 291 Table 7-3 (Continued) PT 9 *C̬ .t10 *C̬ .c11 *C̬ .k12 *C̬ .q13 *k.r14 *p.r15 *t.r16 *c.r17 *q.w18 *m.l19 *C̬ .r20 *C̬ .l21 *t.n22 *C̬ .n23 *g.r24 *q.s25 *m.p26 *s.c27 *z.ɟ28 *r.t29 *t.h30 *n.m31 *gm̩ .r32 *ɟm̩ .r33 *c.pl- Li *tʰ*s*kʰ*kʰ*kʰl*pr*tʰr*tʰl*ʰw*ml/r*dr*dl*ˀdl/r*nl/r*ɣr*kr*p*s*z*tʰr*tʰr*n*br*vr*pl- Sarawit *tʰ*s*kʰ*kʰ- *kʰl-, *xʰrw*pr- Ferlus *nt3- *ŋx3*ŋk3*kʰr1*p.t1- *tʰl- *cʰr*ʰw- *ʰw*ml- *ml*dr- *r/l *dl- *l/r *ˀdr- *ɗ- (<*ˀl-) *nr- *C.n*ɣr- *ɣr*kr*p- *mp1- *br*vr- *mr- 292 The system of PT rimes proposed in this reconstruction also differs markedly from previous reconstructions both with respect to the vowels and the codas. Although the current system agrees with Li’s proposal that PT contrasted three vowel heights, the two systems disagree with the respect to vowel quantity. Unlike Li’s reconstruction, the current system, like Sarawit’s, posits a vowel length contrast for each vowel quality. Consequently, the quality distinction in the conventional system between mid and low vowels is replaced by a length distinction. As for the codas, the current system proposes a richer set of coda consonants that includes *-l, *-c, and *-ɲ in addition to those posited in Li’s reconstruction. Most importantly, this reconstruction crucially differs from Li’s in not reconstructing the typologically suspect contrast between pairs like *wo and *uo, or pairs like *oɰ and *ɤɰ. Table 7-4 shows clearly that the rime system of the current reconstruction is simpler and more symmetrical than those of Li’s and Sarawit’s systems. Table 7-4 PT rimes and their correspondents in previous reconstructions PT 1 *-i:# 2 *-iC 3 *-i:C 4 *-ɯ:# 5 *-ɯC 6 *-ɯ:C 7 *-u:# 8 *-uC 9 *-u:C 10 *-e:# Li -i:# *-iC *-iəC *-ɯ#, *-ɰu# *-wɯn, *-ɯum, *-ɯŋ, *-ej, *-ew *-ɯən, *-ɯun, *-wom, *-ɯut, *-ɰuk *-u# *uC, *-ouk *-uəC *-ɛ# Sarawit *-i:# *-iC *-i:C *-ɯ:# *-ɯC, *-oim, *-ɤŋ, *-ɤw, *-ɤj *-ɯ:C, *-waC *-u:# *-uC *-u:C, *-eu:k, *-eu:# *-e:# 293 Table 7-4 (Continued) PT 11 *eC 12 *-e:C 13 *-ɤ:# 14 *-ɤC 15 *-ɤ:C 16 *-o:# 17 *-oC 18 *-a:# 19 *-aC 20 *-a:C 21 *-iə 22 *-ɯə 23 *-uə 24 *-aɰ Li *-eC, *-jeC *ɛC *-ɯ#, *oC, *-ɛj, *-eɰ/-ɛɰ, *-jɤw *-ɯaC, *-oɰ#, *-ɔ# *-oC, *-woC, *-ow *-a# *-ɤC *-aC *-ie, *-iaw, * -iaj *-ɯe, *-ɯaj, *-uɯ *-ue, *-uaj *-ɤɰ Sarawit *eC , *ei:, *-ɤn *-e:C *-wo: *-ɤC, *-oC, *-ia:j, *-ia:w, *ɤ:w *-ɤ:C, *-ɯa:C, *-ua:ɰ, *ua:# *-o:# *-ok, *-oŋ, *-oit, *-oi:C, *-euC *-a:# *-aC *-a:C *-ia *-ɯa, *-ɯa:C, *-ua:C *-ua *-aɰ Last but not least, the tonal system of PT in the current system is identical to earlier reconstructions in terms of number of categories. In PT, there were four tonal categories—*A, *B, *C, and *D. However, the current reconstruction goes beyond Li and other researchers in proposing a concrete reconstruction of these categories. Pitch, voice quality, and vowel duration are all claimed to be part of the identities of the tones. In particular, PT *A was a modal tone with a mid level contour, *B was a creaky tone with a low rising contour and a relatively long vowel duration, *C was a high falling contour ending in a glottal constriction with a relatively short vowel duration, and *D was a low rising contour restricted to checked syllables. 294 My reconstruction nicely accounts for the three outstanding problems in the historical development of Tai languages discussed in §1.3.2. The first problem is the large number of correspondences not satisfactorily accounted for by earlier reconstructions. This problem is overcome by distinguishing loanwords from native PT words, and by positing sesquisyllabic roots alongside the conventionally reconstructed monosyllabic roots. This is clearly illustrated by the case of the “voicing alternation” discussed in §3.4. The current reconstruction shows that a number of etyma standardly cited as examples of the voicing alternation are in fact not reconstructible at the PT level, e.g. ‘bean’, ‘ten’ etc. For etyma that cannot be explained as borrowings, the current reconstruction proposes that they were derived from sesquisyllabic clusters, e.g. *C̬ .tɤŋA ‘to arrive’. Other Kra-Dai languages show plain voiceless stops for these sesquisyllabic etyma, e.g. Proto-Kam-Sui *taŋA ‘to arrive’, supporting the claim that they go back to clusters containing voiceless stops. The second problem faced by earlier reconstructions, Li’s in particular, is the complexity of the vowel system. My reconstruction approaches this problem by taking seriously the conditioning environments for vowel changes. The “Gedney’s Puzzle” paradigm illustrates clearly how the proposed vocalism handles the problem. Rather than positing complex rimes for etyma in this paradigm as in Li’s reconstruction, the current approach argues that the observed pattern of reflexes resulted from divergent changes of the vowels conditioned by final glides. The vowel *ɯ in *pɯwA ‘crab’ became rounded in Siamese due to the influence of the final glide *-w. The resulting *-uw then merged with PT *-u: to give /pu:A1/ in modern Siamese. The situation is different from Yay, where *-ɯ lowered to *-a probably through dissimilation from *-w. The resulting *-aw then merged with the inherited PT *-aw to give modern Yay /paw A1/. 295 The third problem is the aberrancy of Saek, which shows numerous idiosyncracies irreconcilable within all earlier reconstructions. The current reconstruction views most of these properties as clues to the PT phonology not retained elsewhere. For example, etyma that show reflexes of earlier *w- in Saek but *ɓ- in all other Tai languages are reconstructed with sesquisyllabic clusters in PT, e.g. *C̥ .ba:ŋA ‘thin’ for Saek /va:ŋA1/ (< *wa:ŋA/) but Siamese /ba:ŋA1/ (< *ɓa:ŋA/). Another example is Saek /-l/, which is viewed in this reconstruction as a retention of PT final *-l. It should be noted that not all idiosyncrasies in Saek are retentions from PT. The /j-/ reflex of PT *z- is clearly an innovation on the part of Saek, most likely due to the influence of Vietnamese. Further, Li’s and other previous reconstructions suffer from the abstractness of the reconstructed tones. They do not offer a concrete characterization of the tonal categories *A, *B, *C, and *D. The current reconstruction adopts an explicit framework of tone change and combines internal reconstruction and the Comparative Method to arrive at a concrete reconstruction of PT tones. The proposed tonal reconstruction describes PT tones in terms of pitch level, pitch contour, voice quality, and vowel duration. Typologically, it characterizes PT as a pitch-plus-voice quality tonal system. In sum, the current reconstruction differs markedly from Li’s reconstruction with respect to all aspect of the phonology: structure of prosodic words, onsets, rimes, and tones. It accounts for outstanding problems that cannot be resolved within Li’s system including the large number of onset correspondences, the complexity of the vowel system, the aberrancy of Saek, and the abstractness of the tonal categories. 296 7.3 Tai subgrouping The conventional subgroup structure of the Tai family is that proposed by Li (1960; 1977), as discussed in §1.2. According to this view, the Tai language family comprises three primary branches, Southwestern Tai (SWT), Central Tai (CT), and Northern Tai (NT), as illustrated in Figure 7-1. Alternative views (Chamberlain 1975; Ferlus 1990; Gedney 1989b; Haudricourt 1956; Strecker 1985) hold that CT and SWT form one primary branch which is a sister to NT. Note that Haudricourt (1956) further notes the existence of some understudied dialects that are intermediate between his two primary branches. Figure 7-1 Tai subgrouping according to Li (1960). Although the disagreements among the different proposals are not trivial, it is safe to say that they are all based on the traditional reconstruction of PT, that is, Li (1977). Because the current reconstruction of PT phonology differs markedly from the traditional system, the subgrouping it implies will be significantly different from its 297 predecessors. Using phonological innovations as criteria for subgrouping, I propose in this section a tentative subgrouping of Tai dialects from the perspective of the current reconstruction without discussing earlier proposals in detail. The proposed subgroup structure of Tai is given in Figure 7-2. Figure 7-2 Tentative subgroupings of some Tai dialects studied Proto-Tai AB ru Ningming EF ru Lungchow G H Leiping ru Lungming K L Daxin ru Debao CD Chongzuo ru Shangsi I J Qinzhou ru MN O P Jingxi Wuming Saek ru Bao Yen G. Nung Yongnan Po-ai Q R Cao Bang W. Nung Long’an Yay Shan Sapa Wenma Y. Nung Fusui Lingyue Siamese Rong’an Black Tai Qiubei Lue Bouyei Other SWT dialects Other NT dialects The Tai family tree shown in Figure 7-2 contains altogether 10 terminal nodes: B, C, F, H, I, L, M, N, P, Q, and R. The varieties grouped together under these nodes are combined to form larger subgroups represented in the family tree by nodes A, D, G, I, J, K, and O. Methodologically, I assume that a single sound change does not define a subgroup because most, if not all, innovations can diffuse through contact among dialects. What determines a subgroup is the clustering of several innovations. Furthermore, I assume that innovations that define each subgroup must be exclusively shared only by the members of the subgroup. In fact, many innovations are used as 298 criteria for defining more than one group. Because the proposed classification is preliminary in nature, the precise sets of phonological innovations that define the different subgroups have not been systematically identified. The change from *ɯj and *ɯw to *i: and *u: is a good illustration of how innovations are used in this preliminary subgrouping. This sound change is used as a criterion for both group A and C. For group A, the change from *ɯj and *ɯw to *i: and *u: clusters with four other innovations, three of which are not shared by C. Similarly, subgroup C shows three other innovations, two of which are shared with subgroup D rather than C. Therefore, A and C must be recognized as two separate subgroups. In this sense, Lungchow belongs to subgroup A because it shows clustering of the five innovations that define subgroup A. The membership of the terminal subgroups and the phonological criteria that define them are shown in Table 7-5. Table 7-5 Phonological criteria for subgrouping of Tai languages Groups Innovations A 1) *ɯj, *ɯw = *i:, *u: 2) *ɤj, *ɤw, *ɤɰ = *aj, *aw, *aɰ 3) *we, *wo = *e:, *o: 4) *ɟm̩ .r- > *br- 5) *ɯ:k > *u:k B 1) *ɤj, *ɤw, *ɤɰ = *i:, *u:, *ɯ: 2) *ɯj, *ɯw = *i:, *u: 3) *we:, *wo: = *e:, *o: Varieties Subgroups F, H, L, P, Q, and R Ningming 299 Table 7-5 (Continued) Groups Innovations C 1) *ɯj, *ɯw = *i:, *u: 2) :*we:, *wo: = *e:, *o: 3) *k.t- = *tr- 4) *ɟm̩ .r- > *ɟr- D 1) *ɤj, *ɤw, *ɤɰ = *i:, *u:, *ɯ: 2) *we, *wo = *i:, *u: 3) *k.t- = *tr- 4) *ɟm̩ .r- > *ɟr- E 1) *p.t- = *p.r- 2) *ɯm > *ɤm F G *k.r- = *qr- H *qr- = *ʰr- I *ɤ:# = *a:# J *ɤn, *ɤt, *ɤc = *an, *at, *ac K *e:, *o: > *ɛ: and *ɔ: L *qr- = *kr- M *p.t- = *trN *ɯj, *ɯw = *aj, *aw P Varieties Chongzuo, Shangsi, Caolan Subgroups I, M, and N Subgroups H, L, P, Q, and R Lungchow, Leiping Subgroups L, P, Q, and R Lungming, Daxin Qinzhou Subgroups M an N Subgroups P, Q, and R Debao, Jingxi, Western Nung, Guangnan Nung, Yanshan Nung Wuming, Yongnan, Long’an, Fusui Saek, Yay, Po-ai, Bouyei, Hengxian, Yongbei, Pingguo, Tiandong, Tianlin, Lingyue, Guangnan Sha, Qiubei, Liujiang, Yishan, Huanjiang, Rong’an, Longsheng, Hechi, Nandan, Dong’an, Dulan, Shanglin, Laibin, Guigang, Lianshan. Bao Yen, Cao Bang, Wenma 300 Table 7-5 (Continued) Groups Innovations O *ɤn > *on Q *kr- = *ʰr- R *kr- = *s- Varieties Subgroups Q, and R Black Tai, White Tai, Dehong, Tai Aiton, Tai Phake, Southern Shan, Tai Ya, Lue, Nyo, Yoy, Kaloeng, Phuan, Phu Thai, Lao, Khuen, Tai Yuan, Siamese, Southern Thai, Tak Bai Sapa In Table 7-5 some subgroups are given only one or two innovations, but this does not mean that the subgroups are defined by those innovations only. Varieties belonging to a given subgroup must show all the innovations that are shared by all nodes that dominate that subgroup. For example, Siamese belongs to Q because it shows all the innovations that characterize A, E, G, K, O, and Q. This means that subgroup Q is defined not by one single innovation *kr- > *ʰr- but by 11 innovations. Similarly, Wuming belongs to subgroup M because it shows all innovations associated with nodes D, J, and M. In other words, subgroup M is defined by six innovations altogether. Note that some of the subgroups are defined negatively, In other words, they do not share any innovation that would distinguish them from their sister subgroup(s), but are grouped together due to absence of the innovations that their sisters went through. For example, Bao Yen and Cao Bang are put together in one group because they clearly belong to K but do not show the change from *ɤn to *on, characteristic of subgroup O. To my knowledge, they do not share any innovations that can identify them as members of subgroup O. Alternatively, we might posit a flat structure for 301 subgroup K, which would consist of three separate branches, namely, Cao Bang, Bao Yen, and subgroup O. The proposed preliminary subgrouping of Tai dialects shows some notable features. First, the subgroup represented by node Q comprises languages conventionally included in the reconstructions of PSWT (Jonsson 1991; Li 1977; Pittayaporn 2008b, to appear-b). In this sense, SWT is the only one of Li’s three groups that is recognized as a valid subgroup. It corresponds to subgroup Q in the family tree in Figure 7-2. Note, however that in addition to the dialects previously recognized as SWT included here in the Q group, other varieties sharing a large number of SWT traits are not included in Q. The only such language included in my sample is Sapa. This variety is extremely similar to subgroup Q, but did not go through the change from *kr- to *ʰr- characteristic of SWT. In addition, it shows the change from *kr- to *s- not found in the SWT dialects in subgroup Q. Some understudied Tai varieties along the Red River not included in this study, e.g. Payi (Kullavanijaya & L-Thongkum 2000), and Yuanyang (Zhou & Luo 2001), might have a similar relationship to SWT. Second, none of the subgroups in the proposed subgrouping corresponds exactly to NT in Li’s classification. The subgroup that most closely resembles Li’s NT is subgroup J, which comprises all of Li’s NT dialects plus the CT dialects Yongnan, Long’an, and Fusui in central Guangxi. Of particular relevance is Yongnan, which is almost identical to the NT dialect of Wuming. The only two differences are that in Yongnan *Cr- clusters are reflected as aspirated, and the cluster *tr- that resulted from the merger of PT *tr-, *p.t-, and *p.r-, is reflected as /tʰ-/. In contrast, Wuming has unaspirated reflexes for the *Cr- clusters, and has /r-/ for the intermediate *tr-. Given their closeness to the NT dialect Wuming, the three CT dialects in subgroup J thus constitute a very strong piece of evidence for considering aspiration to be a recent 302 areal development. In short, Li’s NT can be kept as a subgroup only if the defining criteria are modified to include such dialects as Yongnan, Long’an, and Fusui. Third, the CT group does not have any genetic status in this subgrouping schema. Varieties classified as CT by Li’s criteria are distributed in all four primary branches A, B, C, and D in the proposed subgrouping. This supports Strecker’s view (1985) that the phonological innovations shared by CT dialects and used by Li as subgrouping criteria are in fact recent areal innovations. Furthermore, distributing CT dialects among the four primary branches of Tai means that Li’s hypothesized CT by itself in fact encompasses the full genetic diversity of Tai languages. As shown in Figure 7-2, all of Li’s SWT dialects together make up subgroup Q, which is one lowlevel subgroup within a single primary branch (A). Similarly, all NT dialects together with Wuming-like CT dialects form subgroup J, which represents only part of the diversity within primary branch D. This situation straightforwardly captures the homogeneity of SWT and NT widely noted in the literature (Gedney 1989b; Pittayaporn 2008b). Lastly, this proposal does not consider Saek as an outlier of Tai but merely a minor subgroup within one of the primary branches. More specifically, Saek is a member of subgroup N, which comprises all NT dialects in Li’s classification except Wuming. As Saek shares a great number of phonological innovations that characterize NT (Gedney 1989c; Haudricourt 1963a), it follows that Saek belongs to subgroup N within the proposed subgrouping. Therefore, the peculiarities of Saek must be considered relatively recent innovations or retentions from the common ancestor of subgroup N. For example, PT *t.n- must have survived until after the separation of Saek from the rest of subgroup N so that the sesquisyllabic cluster did not undergo the implosivization of *t.n- found in all other Tai dialects. After the separation, Saek went its own way and changed *t.n- into /tr-/. 303 As has become clear, the preliminary subgroup structure proposed here shows characteristics that depart greatly from the conventional three-way classification. It does not recognize either Li’s CT or NT as genealogical subgroups, even though subgroup N resembles NT rather closely. Nor does it view SWT as one of the primary branches. Most importantly, it claims that the Tai family tree is rather flat as it consists of more than two primary branches. In short, the reconstruction of PT phonology proposed in this study implies a subgroup structure radically different from all earlier proposals. 7.4 PT phonology in Kra-Dai context Not only is the current reconstruction of PT phonology consistent with advances in the comparative study of Kra-Dai made in the past decades, but it also presents new ideas that further our understanding of Kra-Dai phonology. In this section, I briefly discuss three implications that the current reconstruction of PT has for Kra-Dai phonology: 1) the shape of the prosodic word, 2) the complexity of the vowel system, and 3) the nature of the lexical contrast underlying the *ABCD system. The first implication is the shape of the prosodic word in Proto-Kra-Dai (PKD). Mounting evidence from other Kra-Dai languages shows that PKD could not have been strictly monosyllabic (see §2.3) but it is not clear exactly what the situation in PKD was. The current reconstruction of PT seems to suggest that the canonical shape of the PKD prosodic word might have been disyllabic. Assuming a genetic relationship between Kra-Dai and Austronesian, Benedict (1982; 1988; 1997a) and Haudricourt (1975) propose that monosyllables in modern Tai languages resulted from a transference of the vowel in the first syllable of disyllables to the second, e.g. PKD *ɓulanA > PT *ɓlɯənA ‘moon’ as discussed in (see §2.3.1). Although I showed that 304 PT was sesquisyllabic rather than disyllabic, such “vocalic transfer”, to use Benedict’s term, is likely to have occurred prior to the formation of PT. At several points, reconstructions in this dissertation seem to support the position that PKD was a disyllabic language. In particular, it suggests that the sesquisyllabic etyma in PT and their cognates in other Kra-Dai languages followed independent but parallel developments from common disyllabic sources. For example, PT *C̥ .dipD ‘raw’ corresponds to Sui (Institute of Language and Culture for Rural Development & Thailand and the Kam-Tai Institute 2003) /ˀdjupDS1/ ‘raw’, suggesting PKD *C̥ u.dipD. The PKD form may have gone through two parallel vocalic transfers, which resulted in the discrepancy in the vowel reflex between the PT and the Sui forms. In Tai, the transferred vowel *u landed in front of the original vowel of the second syllable *i. It was then lost through simplification, i.e. PKD *C̥ u.dipD > *C̥ .duipD > PT *C̥ .dipD. In contrast, in Sui, the transferred vowel *u was relocated after *i, which subsequently underwent through glide formation to *-j-, i.e. *C̥ u.dipD > *C̥ .diupD > Sui /ˀdjupDS1/. If this scenario is correct, PKD must be reconstructed with disyllabic prosodic words. The second implication is the complexity of the vowel system. The current reconstruction presents a very symmetrical and relatively rich vocalism for PT. One of the most daunting challenges in the study of Kra-Dai phonology is its vocalism. The lack of an understanding of the PKD vowel system is reflected in Strecker (1988), who shows that examining the vowel correspondences among Kam-Sui languages alone has proven to be a daunting task. To my knowledge, there has not been any systematic study of the Kra-Dai vocalic system. The literature on comparative Kra-Dai has focused mainly on the consonants, e.g. Ostapirat (2005), and Edmondson and Yang (1988). 305 Putting aside complications created by the vocalic transfer discussed above, a clearer picture of the PKD vowel system promises to emerge as a result of the proposed PT vocalism. The best illustration is perhaps the case of PKD long back unrounded vowels *a:, *ɯ:. In Li’s system, a diverse set of vocalic nuclei including PT *a, *ɯa, *(ɰ)u, and *ɯ, all correspond to Sui (Institute of Language and Culture for Rural Development & Thailand and the Kam-Tai Institute 2003) /a:/87 as shown in Table 7-6. Table 7-6 Proto-Tai and Sui reflexes of PKD back unrounded vowel. Etyma ‘stone mountain’ ‘village’ ‘house’ ‘rope’ ‘ear’ ‘child’ ‘hand’ Li’s *pʰraA *ˀbanC *rɯanA *dʑɯakD *xrɰuA *lɰukD *mwɯA PT *pra:A *ɓa:nC *rɤ:nA *ɟɤ:kD *k.rwɯ:A *lɯ:kD *mwɯ:A Sui pra:A1 ˀba:nC1 ɣa:nA la:kDL1 qʰa:A1 la:kDL2 mja:A2 PKD *a: *ɤ: *ɯ: With the current reconstruction of PT phonology, it has now become clear that Sui /a:/ corresponds regularly to back unrounded *a:, *ɤ:, and *ɯ: in PT as shown in Table 7-6. Most likely, PT retains the PKD long back unrounded vowels intact while Sui merged them all into /a:/. From a Kra-Dai perspective, this systematic correspondence seems to indicate that PKD had a relatively rich vowel system with 87 Sui /a/ in open syllables is retranscribed as /a:/. 306 length contrast. In other words, PKD vocalism may have been very similar to the reconstructed PT vowel system. The last implication to be discussed is the nature of the lexical prosodic contrast in PKD. Although it has never been systematically examined, it is generally assumed that PKD was a tonal language. The basis for this assumption is clearly the typological fact that all attested Kra-Dai languages are tonal. This view is clearly reflected in Ostapirat (2005), who attempts to correlate PAN final consonants to tones *A, *B, *C, and *D in Kra-Dai. Although the current reconstruction of PT phonology does not provide unequivocal evidence for or against this standard view, it highlights the striking similarities among the tonal systems of Kra-Dai languages. The current reconstruction claims that PT had 3 contrastive tones *A, *B, and *C on non-checked syllables but neutralized them into the so-called *D tone on checked syllables. This is exactly the tonal system found in all of the other branches of Kra-Dai, including Kra (Ostapirat 2000), Hlai (Norquest 2007) 88 , and Kam-Sui (Thurgood 1988). Moreover, both pitch and voice quality are claimed to be crucial to the identity of the lexical tones. PT *A was a modal tone with a mid level contour. *B was a creaky tone with a low rising contour. Lastly, *C was a high falling contour ending with a glottal constriction. Clearly not accidentally, the Proto-Kra tonal system as proposed by Ostapirat (2000) is very similar to the proposed PT system with regards to the number of categories and their phonetic contents. This intriguing resemblance among the tonal systems of the different branches needs to be explained. Perhaps the tonal categories were part of PKD, but it is equally plausible that they represent parallel developments in different branches. 88 Norquest (2007) reconstructs Proto-Hlai with an ABCD tonal system but further project these categories back to final laryngeals in the Pre-Hlai stage. 307 The implications discussed above illustrate the relevance of the current reconstruction of PT to the study of Kra-Dai. The proposed reconstructed PT phonology is a significant step forward in understanding the Kra-Dai family as a whole. 7.5 Convergence of Tai varieties The reconstruction of PT phonology proposed in this dissertation presents a clear picture of the development of Tai languages. More specifically, it takes an explicit position that the history of Tai varieties is essentially one of convergence. Despite its ability to explain the range of comparative data, its claim that PT was a sesquisyllabic language is at variance with the typological fact that all modern Tai languages are monosyllabic, at least in the native portion of the lexicon (see §2.3). Although cases in which daughter languages are typologically very different from the parent language are common, the fact that all modern Tai varieties agree in exhibiting strong monosyllabicity, despite the sesquisyllabicity of PT, invites discussion. The situation is quite different from the case of Chamic languages, where the modern languages do not conform to one single type (Lee 1966; Thurgood 1999). Although these mainland Austronesian languages have arguably evolved away from the disyllabicism of Proto-Chamic toward monosyllabicism, they attest various stages of development so that disyllabic, sesquisyllabic, and monosyllabic Chamic languages are all attested. I claim that two factors led to the typological homogeneity of modern Tai dialects: prolonged dialect contact and shared external influence. The first factor is the prolonged contact among Tai dialects. The preliminary subgroup structure reveals a large number of innovations that cut across subgroup boundaries. Such innovations indicate that the dialects remained in contact with each other after initial diversification so that an innovation that originated in one group can diffuse to others. A clear example of an innovation that spread across subgroup 308 boundaries is aspiration of PT *Cr- (see §3.5.2). All modern dialects except for Wuming and those in the N subgroup underwent the process of aspiration but, contra Li (1960), the change does not define any of the subgroups. The case of Yongnan and Wuming discussed above shows clearly that aspiration of *Cr- is a case of diffusion. Another example is the neutralization of *C̬ .l- and *C̬ .r- found commonly in subgroup N. Although the majority of dialects in this subgroup have /r-/ or other consonants that came from earlier *r- (see §3.6.8.1 and §4.4.5) as reflex of this merger, Saek still keeps the two distinct as /l-/ and /r-/ respectively. Similarly, a diverse range of varieties in the A branch show /l-/ for both, but Sapa and the Q subgroup (i.e, subgroup O) still maintain the distinction (see §4.4.5). These two cases are just a few examples of the many phonological innovations that diffuse from one subgroup to others. These innovations are clear attestations of the prolonged dialect contact that led to the observed typological homogeneity among Tai dialects. The second factor is shared external influence. Obviously, the language that had the greatest influence on Tai languages was Chinese (Benedict 1975; Li 1976; Manomaivibool 1975; Maspéro 1912; Wulff 1934). The Chinese influence is evident both in the amount of Chinese vocabulary found in Tai languages and in the fact that the three-plus-one tonal system of PT was structurally identical to the Middle Chinese tonal system. The Chinese influence on Tai must have persisted for an extended period of time as Manomaivibool (1975; 1976) shows successfully that Siamese, despite being the most Indianized of all Tai dialects, nevertheless consists of at least four layers of Chinese loanwords: 1) pre-Middle Chinese (including OC and LH), 2) (Early) Middle Chinese (MC), 3) Late Middle Chinese (LMC), and 4) post-Middle Chinese. Siamese is not the only variety that shows these layers. Etyma from the earliest three layers are found widely among Tai dialects. For example, 稼 jià ‘sow crops’ (OC 309 *krâh, LH kaC, MC kɑC, LMC kja:`) is the source for PT *kla:C ‘rice seedling’. It belongs to the pre-Middle Chinese layer. The presence of the liquid *-l- indicates that the borrowing predated the Later Han dynasty. Similarly, many etyma found widely in Tai belong to the Middle Chinese layer. For instance, Chinese 破 pò ‘to smash’ (OC *phâih, LH phɑiC, MC pʰwâB, LMC pʰua`) is found in Siamese and Lungming as /pʰa:B1/ ‘to split’ and in Yay as /pa:B1/. This post-PT loan must have been incorporated around the Early Middle Chinese period as suggested by the rime /a:/. Lastly, Chinese borrowings from the Late Middle Chinese layer are also commonly found in Tai dialects. For example, 夫 fū ‘husband’ (OC *pa, LH puɑ, MC pju, LMC fuə̆ ) is attested in most if not all SWT and CT dialects, e.g. Siamese /pʰuəA/, and Western Nung /pʰu:A/ ‘husband’. The aspirated /pʰ-/ in modern dialects suggests that OC and MC *p- had already been spirantized to f- in the source language, dating the borrowing of this etymon to Late Middle Chinese. These examples of Chinese loanwords clearly show that Tai languages have been in contact with Chinese since Han times at least, and that contact has been persistent and widespread across Tai varieties, even after differentiation. This prolonged contact with Chinese contributed to the observed typological homogeneity among Tai dialects. The prolonged contact among Tai dialects and the shared Chinese influence were the most important factors responsible for the typological homogeneity of Tai dialects on the one hand, and the typological discrepancy between reconstructed PT and its daughter languages on the other. Even though the sesquisyllabicity of PT as reconstructed here at first glance seems at odds with the monosyllabicity characteristic of modern Tai varieties, such a discrepancy is in fact expected for dialects that remain in close contact for an extended period of time. In this sense, the typological homogeneity among Tai languages is a product of convergence rather than a shared retention from PT. More specifically, Tai dialects, which remained in contact with 310 each other, all followed the same path toward monosyllabicity, with Chinese as a model. As shown by Matisoff (2001), monosyllabicization is clearly an areal phenomenon of Mainland Southeast Asia. In Kra-Dai, this tendency to become monosyllabic is particularly strong, to the extent that all but a very few languages have become monosyllabic. 7.6 Conclusion This dissertation presents a new phonological reconstruction of PT. It takes as its premises a systematic application of the Comparative Method and the force of language contact upon language change. Essentially, the interpretation of the comparative data was constrained by explicit and motivated hypotheses that did not underestimate the role of areal convergence and dialect contact in language change. Moreover, it incorporates a large amount of dialect data, much of which has only become available recently. The resulting reconstruction thus has various advantages over previous proposals. The first advantage is the success of the current reconstruction in accounting for the various problems in historical studies of Tai languages in a unified way. My reconstruction of PT phonology claims that PT was a sesquisyllabic language despite the monosyllabicity characteristic of modern Tai languages. This view of PT connects several phenomena that are generally regarded as unconnected and gives them a unified account. Such phenomena include the voicing alternation, certain alleged split reflexes in Saek, and many other complex correspondences that are viewed here as results of a single phenomenon of monosyllabification. The second advantage of the current reconstruction is its simplicity. Although the proposed reconstructed PT sound system is complex in some areas, such as having sesquisyllablic roots and a richer set of contrasts among places of articulation, it 311 allows us to eliminate many more complexities in other areas of the system. For example, reconstructing a richer place contrast help reduce the number of phonation type contrasts. The establishment of the uvular series and the elimination of the aspiration contrast were independently motivated, but the former got rid of the need to reconstruct aspirated *kʰ-, the only aspirated phoneme in Li’s system that is supported by a large number of cognate sets. Finally, the PT reconstruction here clearly distinguishes between shared retentions and shared innovations that resulted from convergence. One of the most serious shortcomings common among earlier proposals is a lack of appreciation of the effect of language contact. As a result, many phonological changes that diffused among Tai languages through contact were taken to be properties retained from PT. One of the best examples is the case of aspiration. Because contrastive aspiration is one of the most common properties shared by a large set of modern Tai varieties, it is assumed in Li’s reconstruction to have to been a characteristic of PT. The current reconstruction systematically examined Tai-internal comparative data and showed that aspiration developed secondarily in different dialects that remained in contact after PT broke up into different branches. In conclusion, this dissertation is an attempt to overcome the challenging task of distinguishing universal tendencies, chance, diffusion, genetic retentions, and parallel development in the Tai language family of Mainland Southeast Asia, one of the world’s most intriguing linguistic areas. Equipped with the classic Comparative Method, an abundance of highquality comparative data, and an appreciation of language contact, the current reconstruction of PT phonology presents a vivid picture of the history of Tai languages as characterized by divergent changes overridden by sucessive waves of convergent trends that radically transformed the ancestral Proto-Tai language into a network of typologically homogenous dialects that differ markedly from their parent. 312 APPENDIX A LIST OF LANGUAGES 313 Varieties 1 Aiton 2 Bac Va 3 Bao Yen 4 Black Thai 5 Bouyei Alternate names/spellings Tai Aiton Tai Dam, Song Buyi (布衣) Groups Locations Sources SWT Northeastern India Morey (2005) CT Northeastern Vietnam Hudak (1997) Hudak (2008) CT Northwestern Vietnam Pittayaporn (2007c) SWT Northwestern Vietnam, relocated groups Laos, Thailand, and USA) Gedney (n.d.) Hudak (1994) Hudak (2008) NT Southern Guizhou Chinese Academy of Sciences (1959) Institute of Language and Culture for Rural Development, Mahidol University Kam-Tai Institute, Central University for Nationalities (2001) 314 Varieties 6 Cao Bang Alternate names/spellings Groups Locations CT Northwestern Vietnam 7 Chongzuo (崇左) 8 Daxin (大新) 9 Debao (德保) CT Southwestern Guangxi CT Southwestern Guangxi CT Southwestern Guangxi 10 Dehong (德宏) Tai Nüa, Tai Mao, Dai dialects of Mengshi (芒市) and Menglian (孟连) belong to this group. SWT Southwestern Yunnan, northwestern Myanmar 11 Donglan (东兰) 12 Du'an (都安) 13 Fusui (扶绥) NT Northwestern Guangxi NT Northwestern Guangxi CT Central Guangxi Sources Hoang (1997) Hudak (1997) Pittayaporn (2007d) Zhang et al. (1999) Zhang et al. (1999) Kullavanijaya (1992a) Zhang et al. (1999) Harris (1975) Kullavanijaya (2001) Luo (1999) Zhou, Fang, and Xi (2001) Zhang et al. (1999) Zhang et al. (1999) Zhang et al. (1999) Varieties Alternate names/spellings Groups Locations Sources 14 Guangnan Nung Guangnan Nong (广南侬) CT Eastern Yunnan Zhang et al. (1999) 15 Guangnan Sha (广南沙) NT Eastern Yunnan Zhang et al. (1999) 16 Guigang (贵港) NT Central Guangxi Zhang et al. (1999) 17 Hechi (河池) NT Northeastern Guangxi Zhang et al. (1999) 18 Hengxian (横县) NT Central Guangxi Zhang et al. (1999) 19 Huanjiang (环江) NT Northeastern Guangxi Zhang et al. (1999) 315 20 Jingxi (靖西) CT Southwestern Guangxi Zhang et al. (1999) 21 Kaloeng Kaloeng SWT Southern Laos, relocated groups in central and northeastern Thailand Thisapong (1985) 22 Khuen Khün SWT Eastern Myanmar, Northwestern Thailand Hudak (1994) Petsuk (1978) 316 Varieties 23 Laibin (来宾) 24 Lao 25 Lei Ping 26 Lianshan (连山) 27 Lingyue (凌乐) 28 Liujiang (柳江) 29 Long'an (隆安) 30 Longsheng (龙胜) 31 Lue Alternate names/spellings Groups Locations Sources NT Central Guangxi Zhang et al. (1999) Northeast Thai, Isan Leiping (雷平) SWT CT NT NT Laos, northeastern Thailand Southwestern Guangxi Northwestern Guangdong Northwestern Guangxi Gedney (n.d.) Hudak (1997) Hudak (2008) Gedney (n.d.) Hudak (1995) Hudak (2008) Zhang et al. (1999) Zhang et al. (1999) NT Northeastern Guangxi Zhang et al. (1999) CT Central Guangxi Zhang et al. (1999) NT Northeastern Guangxi Zhang et al. (1999) Lü, Tai Lue, Sipsongpanna SWT Northern Thailand, Gedney (n.d.) Tai. Tai Yong is a dialect nortern Laos, Varieties Alternate names/spellings of Lue, cf. Gedney's Lue dialect of Muang Yong. Dai dialect of Jinghong (景 洪 = Chiang Rung) is a Lue dialect. Groups Locations Northwestern Vietnam, Southern Yunnan Sources Hudak (1996) Hudak (2008) Zhou and Luo (2001) 32 Lungchow Longzhou (龙州), Nung Chao CT Southwestern Guangxi Gedney (n.d.) Hudak (2008) Li (1940) Li (1977) Zhang et al. (1999) 317 33 Lungming Longming (龙明) CT Southwestern Guangxi Gedney (n.d.) Hudak (1991a) Hudak (1995) Hudak (2008) 34 Nandan (南丹) NT Northeastern Guangxi Zhang et al. (1999) 35 Ningming (宁明) Ning Ming CT Southwestern Guangxi Hudak (1995) Hudak (2008) Zhang et al. (1999) Varieties Alternate names/spellings Groups Locations Sources 36 Nung Phan Slinh Nung Fan Slihng CT Northeastern Vietnam Freiberger (1964) Freiberger (1976) Saul and Freiberger Wilson (1980) 37 Nyo Nyaw SWT Southern Laos, relocated groups in central and northeastern Thailand Boonsner (1984) 38 Phake Tai Phake SWT Northeastern India Morey (2005) 318 39 Phu Thai Thai Kapong is a dialect of Phu Thai. SWT Southern Laos, relocated groups in central and northeastern Thailand Phothiban and Trongdi (1998) Pittayaporn (2005b) 40 Phuan Thai Phuan SWT Eastern Laos, relocated groups in central and northeastern Thailand Phothiban and Trongdi (1998) Trongrat (1998) 41 Ping Siang Pingxiang (凭祥) CT Southwestern Guangxi Hudak (1995) Hudak (2008) 319 Varieties 42 Pingguo (平果) 43 Po-ai 44 Qinzhou (钦州) 45 Qiubei (丘北) 46 Rong'an (融安) 47 Saek 48 Sapa 49 Shanglin (上林) 50 Shangsi (上思) Alternate names/spellings Groups Locations NT Central Guangxi Bo'ai (剥隘), Jui NT Eastern Yunnan CT Southeastern Guangxi NT Eastern Yunnan NT Northeastern Guangxi NT Northeastern Thailand, Central Laos SWT NT CT Northwestern Vietnam Northwestern Guangxi Southeastern Guangxi Sources Zhang et al. (1999) Hudak (2008) Li (1977) Zhang et al. (1999) Zhang et al. (1999) Zhang et al. (1999) Gedney (n.d.) Gedney (1989c) Hudak (1993) Hudak (2008) Pittayaporn (2007e) Zhang et al. (1999) Pittayaporn (2008c) 320 Varieties 51 Siamese 52 Southern Shan 53 Southern Thai 54 Tai Ya 55 Tai Yuan 56 Tak Bai Alternate names/spellings Groups Locations Sources Zhang et al. (1999) Thai, Standard, Thai, Bangkok Thai SWT Central Thailand Gedney (n.d.) Hudak (2008) Li (1977) The Royal Institute (2003) Shan, Tai Long, Tai Luang, Tai Yai SWT Eastern Myanmar, Northwestern Thailand Cushing (1914) Gedney (n.d.) Hudak (1994) Hudak (2008) Sarawit (1979) Southern Siamese, Pak Tai, Kedah Thai SWT Southern Thailand, Phothiban and Trongdi (1998) northern Malaysia Umar (2003) SWT Eastern Yunnan Xing (2000) Northern Tai, Lanna Tai, Kam Muang SWT Northern Thailand, relocated groups in Central Thailand Hudak (1997) Hudak (2008) Rungruengsri (1991) Kelantan Thai, Jehe SWT Southern Thailand, Phothiban and Trongdi (1998) northern Malaysia 321 Varieties 57 Tiandong (田东) 58 Tianlin (田林) 59 Wenma (文马) 60 Western Nung Alternate names/spellings Wenma Thu (土) Nung Inh Groups Locations NT Northwestern Guangxi NT Northwestern Guangxi CT Eastern Yunnan CT Northwestern Vietnam 61 White Tai Tai Don, Dialects of Jinping (金平) is a dialect of White Tai SWT Northwestern Vietnam, Southwestern Yunnan 62 Wuming (武鸣) NT Central Guangxi Sources Zhang et al. (1999) Zhang et al. (1999) Zhang et al. (1999) Gedney (n.d.) Hudak (2008) Gedney (n.d.) Hudak (1994) Hudak (2001) Hudak (2008) Zhou and Luo (2001) Gedney (n.d.) Hudak (2008) Li (1956) Li (1977) Zhang et al. (1999) 322 Varieties 63 Yanshan Nung 64 Yay 65 Yishan (宜山) 66 Yongbei (邕北) 67 Yongnan (邕南) 68 Yooy Alternate names/spellings Yanshan Nong (侬) Giay Groups Locations Sources CT Eastern Yunnan Zhang et al. (1999) CT Northwestern Vietnam NT Northeastern Guangxi NT Central Guangxi Gedney (n.d.) Hudak (1991b) Hudak (2008) Pittayaporn (2006) Zhang et al. (1999) Zhang et al. (1999) CT Central Guangxi Zhang et al. (1999) SWT Southern Laos, relocated groups in central and northeastern Thailand Boonsner (1984) 323 Gloss A. Body parts 1 head (1) 2 head (2) 3 head hair 4 hair knot 5 body hair, feather 6 gray-haired 7 forehead 8 nose 9 mucus of the nose 10 face 11 eye 12 mouth 13 tongue 14 tooth 15 gum; gill 16 saliva APPENDIX B GLOSSARY OF RECONSTRUCTED PT ETYMA PT Siamese Sapa Bao Yen Cao Bang Lungchow Shangsi Yay Saek *krawC *truəA *prɤmA *klawC *q.pɯlA *ʰŋwu:kD *pra:kD *ɗaŋA *mu:kD *ʰna:C *p.ta:A *pa:kD *li:nC *wanA *ʰŋɯəkD *la:jA huəA1 pʰomA1 kla:wC1 kʰonA1 ŋɔ:kDL1 pʰa:kDL1 ʔdaŋC1 -t mu:kDL2 na:C1 ta:A1 pa:kDL1 linC2 fanA2 ŋɯəkDL1 la:jA2 hu1 pʰumA1 hoʔDL1 daŋA1 muʔDL2 na:C1 ta:A1 paʔDL1 linC2 hɯʔDL1 huəA1 pʰjomA1 cawC1 kʰonA1 hɔkDL1 -v pʰja:kDL1 daŋA1 mu:kDL2 na:C1 pʰja:A1 pa:kDL1 li:nC2 pʰanA2 hɯəkDL1 la:jA2 tʰuəA1 pʰjomA1 cawC1 kʰonA1 pʰja:kDL1 daŋA1 na:C1 tʰa:A1 pa:kDL1 linC2 vanA2 hɯəkDL1 na:jA2 -i hu:A1 pʰjumA1 kjawC1 kʰunA1 pʰja:kDL1 daŋA1 mukDL2 na:C1 ha:A1 pa:kDL1 linC2 fanA2 hɤ:kDL1 la:jA2 lawC1 pʰomA1 pʰakDL1 daŋA1 mukDL2 na:C1 tʰa:A1 pa:kDL1 lɤnC2 ma:jA2 cawC1 tʰrawC1 piəmA1 pʰramA1 punA1 pja:kDL1 daŋA1 mukDS2 naC1 taA1 pa:kDL1 linC2 fanA2 ŋɯkDS1 na:jA2 pulA1 hu:kDL1 pʰra:kDL1 daŋA1 mu:kDL2 na:C1 pra:A1 pa:kDL1 li:nC2 mla:jA2 324 Gloss 17 cheek 18 ear 19 fang 20 chin, jaw 21 beard 22 neck 23 goiter 24 shoulder 25 arm 26 elbow 27 hand 28 finger, toe 29 fingernail, toenail 30 joint 31 leg 32 knee 33 shin, lower leg 34 foot 35 heel 36 chest 37 back PT *ke:mC *krwɯ:A *χe:wC *ɢa:ŋA *momB *ɣo:A *ʰniəŋA *C̥ .ba:B *qe:nA *C̬ .swo:k D *mwɯ:A *ni:wC *C̬ .lepD *χo:C *p.qa:A *χowB *ɣe:ŋB *ti:nA *sɤnC *ʔɤkD *ʰlaŋA Siamese kɛ:mC1 hu:A1 kʰiəwC1 kʰa:ŋA2 kʰɔ:A2 niəŋA2 ba:B1 kʰɛ:nA1 sɔ:kDL1 mɯ:A2 niwC2 lepDS2 kʰɔ:C1 kʰa:A1 kʰawB1 kʰɛŋB2 ti:nA1 sonC1 ʔokDS1 laŋA1 Sapa kemC1 suA1 xewC1 ka:ŋA2 xoA2 niəŋA1 ba:B1 xenA1 soʔDL1 mɯA2 niwC2 lipDS2 xoC1 xa:A1 xawB1 xeŋC1 tinA1 sunC1 ʔɯkDS1 Bao Yen kɛ:mC1 su:A1 kʰɛ:wC1 ka:ŋA2 kʰɔ:A2 niəŋA1 ba:B1 kʰɛ:nA1 tʰɔ:kDL1 mɯ:A2 niwC2 lopDS2 kʰɔ:C1 kʰa:A1 kʰawB1 kʰɛ:ŋB2 ti:nA1 tʰɤnC1 laŋA1 Cao Bang kɛmC1 suA1 kʰɛwC1 g̈ a:ŋA2 mumB2 gɔA2 niəŋA1 ba:B1 kʰɛnA1 Lungchow ke:mC1 hu:A1 kʰe:wC1 ka:ŋA2 mumB2 ko:C1 ni:ŋA1 ba:B1 kʰe:nA1 Shangsi lyA2 hewC1 kʰa:ŋA2 -i mumB2 hoA2 ba:B1 kenA1 ɬɔkDL1 mɯA2 niwC2 lepDS2 kʰɔ:C1 kʰa:A1 kʰawC1 g̈ ɛŋB2 tinA1 ɬɤnC1 ɁɤkDS1 laŋA1 mɯ:A2 ni:wC2 lipDS2 kʰa:A1 kʰawB1 ke:ŋB2 ɬɤnC1 ʔɤkDS1 laŋA1 moyA2 lipDS2 ha:A1 høA1 hɤŋB1 tɤnA1 ɬonC1 ɁakDS1 laŋA1 Yay cemC1 rɯəA2 -t hewC1 ha:ŋA2 mumB2 hoA2 baB1 cenA1 suəkDL2 fɯŋA2 ritDS2 hoB2 -t kaA1 hoB1 heŋB2 tinA1 θanB1 -t ʔakDS1 laŋA1 Saek ke:mC1 ruəA2 -t ɣa:ŋA2 ɣɔ:A2 va:B1 ke:nA1 suəkDL2 mɯ:A2 li:pDL2 kwa:A1 kɔ:C1 -i ɣe:ŋB2 ti:nA1 sɔnC1 laŋA1 325 Gloss 38 waist (1) 39 waist (2) 40 stomach, belly 41 navel (1) 42 navel (2) 43 buttock 44 pubic hair 45 vagina 46 penis 47 testicles 48 pubic mound 49 urine 50 excrement 51 fart 52 bone 53 bone marrow 54 tendon 55 side, ribs 56 blood 57 meat, flesh 58 skin PT *ˀje:wA *C̥ .wɯǝtD *dwu:ŋC *ɗwɯ:A *ɓli:C *konC *ʰmwu:jA *hi:A *ɣwajA *tram *ʰnawB *niəwB *C̬ .qɯjC *k.tɤtD *C̥ .dukD *ʔwu:kD *ˀjenA *k.ra:ŋC *lɯətD *n.mɤ:C *ʰnaŋA Siamese ʔewA1 tʰɔ:ŋC2 dɯ:A1 konC1 mɔjA1 hi:A2 kʰuəjA2 hamA1 nawB1 jiəwB2 kʰi:C1 totDS1 du:kDL1 ʔenA1 kʰa:ŋC1 lɯətDL2 nɯǝC2 naŋA1 Sapa ʔiwA1 toŋC2 dɯA1 kunC1 mojA1 hiA1 xwajA2 hamA1 newB2 tutDS1 dukDS1 ʔinA1 sa:ŋC1 lɯtDL2 Bao Yen ʔe:wA1 tɔ:ŋC2 konC1 mɔ:jA1 hi:A1 vɤjA2 hamA1 Cao Bang d̤ oŋC2 dɯəA1 diC1 konC1 mɔjA1 hiA1 wɤjA2 tʰamA1 Lungchow ji:wA1 to:ŋC2 kunC1 hi:A1 Shangsi hutDL1 toŋC2 boyC1 -t moyA1 hoyA1 wajA2 tʰamA1 nɛ:wB2 kʰi:C1 tɤtDS1 du:kDL1 ʔɔ:kDL1 ʔɤnA1 sa:ŋC1 lɯətDL2 nɯəC2 naŋA1 nɛwB2 kʰiC1 tɤtDS1 dukDL1 ɁɔkDL1 ɁiənA1 ɬa:ŋC1 -i lɯətDL2 nangA1 kʰi:C1 dukDL1 ʔukDL1 jinA2 ha:ŋC1 lɤ:tDL2 nɯ:C2 naŋA1 newB2 kʰoyC1 tʰotDS1 dokDL1 ɁukDS1 ɲenA1 lutDL2 naŋA1 Yay ʔewA1 hɯǝtDL1 tuŋC2 dɯəA1 miA1 ramA1 nawB1 ɲuB2 hajC2 ratDS1 dokDL1 ʔukDS1 ɲinA2 lɯətDL2 noB2 naŋA1 Saek ʔɛ:wA1 vuǝtDS1 tʰuŋC2 duəA1 ramA1 nawB1 ɲu:B2 ɣajC2 rɛtDS1 rɔ:kDL1 ʔu:kDL1 ʔinA1 luətDL2 mlɔ:B2 naŋA1 326 Gloss 59 sweat 60 pus 61 impetigo 62 wound 63 scurf 64 brain 65 heart, breath 66 lung 67 intestine 68 liver 69 spleen 70 gall bladder, bile 71 illness, fever 72 cold (n.) 73 epidemic 74 kidney-stones B. Animals 75 dog 76 pig 77 mouse, rat PT *r.tɯəB *ʰno:ŋA *kritD *ɓa:tD *ɣɯjA *ʔe:kD *cɤɰA *pwɤtD *sajC *tapD *ma:mC *ɓli:A *k.raj *q.watD *ra:B *ʰni:wC Siamese hɯəB1 nɔ:ŋA1 hitDS1 ba:tDL1 Sapa hɯB1 noŋA1 ba:tDL1 cajA pɔ:tDL1 sajC1 tapDS1 ma:mC2 di:A1 kʰaj1 watDS1 ha:B1 -i niwC1 ʔeʔDL1 caɰA1 sajC1 tapDS1 biA1 sajC1 Bao Yen hɯəB1 nɔ:ŋA1 kʰitDS1 ba:tDL1 kʰɤjA2 cɤɰA1 pɔ:tDL1 tʰɤjC1 tapDS1 ma:mC2 bi:A1 kʰɤjC1 Cao Bang tʰɯəB1 nɔŋA1 kʰitDS1 ba:tDL2 g̈ ɤjA2 cɤɰA1 putDS1 ɬɤjC1 tapDS1 ma:mC2 diA1 -i kʰɤjC1 watDS1 ra:B2 niwC1 Lungchow hɯ:B1 no:ŋA1 ba:t5 ʔe:kDL1 caɰA1 pɤtDS1 ɬajC1 tapDS1 di:A1 hajC1 Shangsi tyB2 nøŋA1 litDS1 hoyA2 coyA1 putDL1 ɬajC1 tapDS1 ma:mC2 boyA1 la:C1 -t Yay noŋA1 ba:tDL1 hiA2 sɯA1 putDS1 θajC1 tapDS1 ma:mC2 diA1 cajC1 niwC1 Saek nɔ:ŋA1 hitDS1 ba:tDL1 ɣi:A2 cɯ:A1 sajC1 tapDS1 ma:mC2 bli:A1 vatDS1 ha:B1 ni:wC1 *ʰma:A *ʰmu:A *ʰnu:A ma:A1 mu:A1 nu:A1 ma:A1 muA1 nuA1 ma:A1 mu:A1 nu:A1 ma:A1 muA1 nuA1 ma:A1 mu:A1 nu:A1 ma:A1 mawA1 nawA1 maA1 muA1 nuA1 ma:A1 mu:A1 nu:A1 327 Gloss 78 water buffalo 79 horse 80 bear 81 jackal 82 elephant 83 deer 84 antelope 85 monkey 86 ape 87 flying squirrel 88 squirrel 89 porcupine 90 civet cat 91 otter 92 pangolin 93 chicken 94 young hen 95 duck 96 goose 97 bird 98 crow PT *ɣwa:jA *ma:C *ʰmwɯjA *nwajA *ɟa:ŋC *kwɯəŋA *wa:lA *li:ŋA *kaŋA *ba:ŋB *ro:kD *ʰmenC *ʰɲelA *na:kD *lilB *kajB *C̬ .qɤ:ŋB *pitD *ha:nB *C̬ .nokD *ka:A Siamese kʰwa:jA2 ma:C2 mi:A1 najA2 cʰa:ŋC2 kwa:ŋA1 fa:nA2 liŋA2 kaŋA1 ba:ŋB1 rɔ:kDL2 menC1 henA1 na:kDL2 kajB1 kʰɯǝŋB1 petDS1 -v ha:nB1 nokDS2 ka:A1 Sapa xwa:jA2 ma:C2 miA1 ca:ŋC2 kwa:ŋA1 fa:nA2 liŋA2 kaŋA1 ba:ŋB1 hoʔDL2 hinA1 naʔDL2 kajB1 xɯəŋB1 pitDS1 hanB1 nukDS2 ka:A1 Bao Yen wa:jA2 ma:C2 mi:A1 nwɤjA2 ca:ŋC2 kwa:ŋA1 liŋA2 kaŋA1 ba:ŋB1 rɔ:kDL2 mɤnC1 hɤnA1 na:kDL2 kajB1 -v kʰɯəŋB1 pɤtDS1 -v ha:nB1 nokDS2 ka:A1 Cao Bang wa:jA2 ma:C2 miA1 nwɤjA2 za:ŋC2 kwa:ŋA1 liŋA2 kaŋA1 ba:ŋB1 Lungchow va:jA2 ma:C2 mi:A1 ca:ŋC2 kwa:ŋA1 liŋA2 kaŋA1 Shangsi wa:jA2 ma:C2 muyA1 najA1 fa:nA2 kaŋA1 menC1 henA1 na:kDL2 linB2 kajB1 -v minC1 hinA1 kajB1 petDS1 -v ha:nB1 nokDS2 ka:A1 pitDS1 nukDS2 ka:A1 mɤnC1 na:kDL2 lɤnB2 kajB1 kʰyŋB1 pitDS1 ka:A1 Yay va:jA2 maC2 mɯəjA1 najA2 sa:ŋC2 vɯəŋA2 fa:nA2 liŋA2 kaŋA1 ba:ŋB1 rokDL2 ɲanA1 na:kDL2 linB2 kajB1 pitDS1 ha:nB1 rokDS2 Saek va:jA2 ma:C2 mɯəjA1 najA2 sa:ŋC2 vuəŋA2 va:lA2 li:ŋA2 rɔ:kDL2 manC1 ɲɛlA1 na:kDL2 lilB1 kajB1 pitDS1 ha:nB1 nɔkDS2 ka:A1 328 Gloss 99 swallow (n.) 100 owl 101 hawk 102 fish 103 catfish 104 shellfish 105 crab 106 shrimp 107 small shrimp 108 frog 109 small frog 110 tree frog 111 turtle 112 water tortoise 113 snake 114 crocodile 115 land leech 116 aquatic leech 117 bedbug 118 grasshopper 119 mosquito PT *ʔe:nB *gawC *lamB *pla:A *C̥ .dokD *ho:jA *pɯwA *kuŋB/C *ŋiəwA *kɤpD *krwe:C *pa:tD *tawB *ʰwɯəA *ŋwɯ:A *ŋɯəkD *da:kD *pli:ŋA *C̬ .rɯətD *p.takD *ɲuŋA Siamese ʔɛnB1 kʰawC2 pla:A1 dukDS1 hɔjA1 pu:A1 kuŋC1 kopDS1 pa:tDL1 tawB1 fa:A1 ŋu:A2 ŋɯəkDL2 tʰa:kDL2 pliŋA1 rɯətDL2 takDS2 juŋA2 Sapa ʔenC1 kawC2 pa:A1 dukDS1 hojA1 puA1 kuŋC1 kupDS1 pa:tDL1 tawB1 fa:A1 ŋuA2 ŋɯʔDL2 taʔDL2 piŋA1 takDS1 ɲuŋA2 Bao Yen ʔɛ:nB1 kawC2 lamB2 pja:A1 dukDS1 hɔ:jA1 pu:A1 kuŋC1 kopDS1 Cao Bang ɁɛnB1 g̈ awC2 -v lamB2 pja:A1 dukDS1 hɔjA1 puA1 kuŋC1 kopDS1 Lungchow ʔe:nB1 kawC2 lamB2 pja:A1 ho:jA1 pu:A1 kuŋC1 ji:wA2 kupDS1 Shangsi ɁenB1 kʰeŋA2 pa:A1 dokDS1 pawA1 koŋB ŋi:wA2 kopDS1 pa:tDL1 tɤwB1 pʰa:A1 ŋu:A2 ŋɯəkDL2 ta:kDL2 piŋA1 rɯətDL2 ɲuŋA2 pa:tDL1 tɤwB1 pʰa:A1 ŋuA2 d̤ a:kDL2 piŋA1 lɯətDL2 tʰakDS1 ɲuŋA2 ŋu:A2 ŋɤ:kDL2 tekDL2 -v piŋA1 lɤ:tDL2 juŋA2 tawB1 pʰyA1 ŋyA2 ŋykDL2 pɤŋA1 lutDL2 tʰakDS1 ɲoŋA2 Yay ʔenB1 pjaA1 dukDS1 pawA1 kuŋB1 kapDS1 kweC1 pa:tDL1 ŋɯəA2 ŋɯəkDL2 ta:kDL1 piŋA1 takDS1 ɲuŋA2 Saek ʔe:nB1 lamB2 pla:A1 rɔkDS1 -v pawA1 kuŋC2 kapDS1 tʰrɛ:C1 viəC1 ŋuəA2 ŋuəkDL2 tʰa:kDL2 pi:ŋA1 ruətDL2 ɲuŋA2 329 Gloss 120 wasp 121 bee 122 insect 123 ant 124 termite 125 worm 126 earthworm 127 gadfly 128 head louse 129 tick 130 flea 131 body louse 132 chicken louse 133 spider 134 coconut grub 135 stink bug 136 caterpillar 137 silkworm 138 butterfly 139 centipede 140 cockroach PT *b.twi:lA *to:B *m.le:ŋA *mɤcD *mo:tD *ʰno:lA *t.nɯəlA *ʰlɯəkD *trawA *trepD *ʰmatD *m.lelA *rwɤjA *krwa:wA *ɗuəŋC *ge:ŋ1 *C̥ .boŋC *mo:nC *ɓɤ:j *q.sipD *sa:pD Siamese tɛ:nA2 tɔ:B1 lɛ:ŋA2 motDS2 mɔ:tDL2 nɔ:nA1 dɯənA1 lɯəpDL1 -f hawA1 hepDS1 matDS1 lenA2 rajA2 duəŋC1 kʰɛ:ŋA2 buŋC1 kʰepDS1 sa:pDL1 Sapa tenA1 toB1 meŋA2 mutDS2 nonA1 dɯənA1 hawA1 hipDS1 matDS1 sawC1 duəŋC1 buŋC1 monC2 bɯC1 xipDS1 Bao Yen pʰɛ:nA1 tɔ:B1 mje:ŋA2 motDS2 mɔ:tDL2 lɯəkDL1 hɤwA1 matDS1 mɤnA2 sa:wA1 duəŋC1 kɛ:ŋA2 boŋC1 mɔ:nC2 bɯəC1 kʰopDS1 Cao Bang tʰɛnA1 tɔB1 mɛŋA2 mɤtDS2 mɔtDL2 nɔnA1 dɯənA1 lɯəpDL1 -f tʰɤwA1 tʰepDS1 matDS1 menA2 rɤjA2 sa:wA1 duəŋC1 g̈ ɛŋA2 mɔnC2 bɯəC1 kʰepDS1 ɬa:pDL1 Lungchow pʰe:nA1 to:B1 me:ŋA2 mɤtDS2 no:nA1 dɤ:nA1 hawA1 matDS1 minA2 ɬajA2 kʰja:wA1 ke:ŋA2 mu:nC2 kʰipDS1 ɬa:pDL1 Shangsi tinA2 tøB1 meŋA2 motDS2 motDL2 nønA1 nunA1 tʰawA1 tʰitDS1 -f matDS1 mɤnA1 loyA2 ha:wA1 -i boŋC1 litDS1 -f ɬa:pDL1 Yay tinA2 neŋA2 matDS2 motDL2 nonA1 dɯənA1 lɯəkDL1 rawA1 matDS1 nanA2 riA2 ca:wA1 duŋC2 keŋA2 ba:C2 θipDS1 Saek tʰi:lA2 tɔ:B1 mɛ:ŋA2 mɛkDS2 mɔ:tDL2 nɔ:lA1 truəlA1 rawA1 ripDS1 matDS1 mlɛlA2 ri:A2 tʰra:wA1 duəŋC1 vɔŋC2 ba:C1 sa:pDL1 Gloss 141 horn 142 tusk, ivory 143 wing (1) 144 wing (2) 145 tail 146 fish scales 147 fishbone 148 egg 149 gizzard 150 cockscomb 151 cockspur 152 hump of an ox 153 stinger (of a bee) PT *qawA *ŋa:A *wɯətD *pi:kD *trwɤ:ŋA *klecC *ka:ŋC *qraj *p.taɰA *ho:nA *t.nɤ:jA *ʰno:kD *lajA Siamese kʰawA1 ŋa:A2 Sapa ŋa:A2 pi:kDL1 ha:ŋA1 klet2 ka:ŋC1 kʰaj2 tajA1 ŋɔ:nA1 dɯəjA1 nɔ:kDL1 najA2 -i ha:ŋA1 kitA2 ka:ŋC1 sajB1 taɰA1 honA1 dɯA1 lajA2 Bao Yen sawA1 ŋa:A2 Cao Bang ɲa:A2 Lungchow Shangsi ŋa:A2 pi:kDL1 ha:ŋA1 kɤtA2 kʰajB1 pʰɤɰA1 hɔ:nA1 dɯəA1 pikDL2 kepDS1 -f ka:ŋC1 sajB1 -v tʰɤjA1 hɔnA1 dɯəA1 nɔkDL1 lɤjA2 pikDL1 ha:ŋA1 kitB1 kʰjajB1 ho:nA1 pikDL1 tʰyŋA1 kipDS1 -f ka:ŋC1 lajB1 toyA1 hønA1 Yay kawA1 ŋaA2 fɯətDL2 rɯəŋA1 cipC1 -f cajB1 taɰA1 daA1 nokDL1 lajA2 Saek kawA1 ŋa:A2 viətDL2 ruəŋA1 tlɛkA2 ka:ŋC1 tɤ:A1 hɔ:nA1 tra:A1 nɔ:kDL1 lajA2 330 C. Plants 154 tree, wood 155 leaf 156 big leaf for wrapping 157 flower 158 banana blossom 159 stem *mwajC *ɓaɰA *k.to:ŋA *ɓlo:kD *pli:A *ka:nC ma:jC2 bajA1 tɔ:ŋA1 dɔ:kDL1 pli:A1 ka:nC1 majC2 baɰA1 toŋA1 boʔDL1 piA1 ka:nC1 majC2 -v bɤɰA1 tɔ:ŋA1 bjɔ:kDL1 pi:A1 maj C2 -v bɤɰA1 tɔŋA1 bjɔkDL1 majC2 baɰA1 to:ŋA1 bjo:k5 pi:A1 majC2 boyA1 tʰøŋA1 bokDL1 majC2 -i baɰA1 roŋA1 dokDL1 ka:nC1 majC2 bɤ:A1 rɔ:ŋA1 blɔ:kDL1 pli:A1 ka:nC1 331 Gloss 160 peel, bark 161 husk 162 thorn 163 fruit 164 sheath, pod 165 grain 166 stump 167 root 168 clump (as of bamboo) 169 sprout, shoot 170 bamboo shoot 171 ear (of rice) 172 rice 173 rice seedling 174 husked rice 175 millet 176 resin 177 banana 178 sugarcane 179 bamboo 180 cucumber, melon PT Siamese *plɯəkD plɯəkDL1 *ka:pD *ʰna:mA *ʰma:kD *q.wakD *m.lecD *to:A *C̬ .ra:kD *ko:A *ʰno:B *r.na:ŋA *rwɯ:ŋA *C̬ .qawC *kla:C *sa:lA *ʰwɯǝŋC *ˀja:ŋA *kluəjC *ʔo:jC *prajB *p.rwe:ŋA ka:pDL1 na:mA1 ma:kDL1 fakDS1 metDS2 tɔ:A1 ra:kDL2 kɔ:A1 nɔ:B1 ruəŋA2 kʰa:wC1 kla:C1 sa:nA2 fa:ŋC1 ja:ŋA1 kluəjC1 ʔɔjC1 pʰajB1 tɛ:ŋA1 Sapa pɯʔDL1 ka:pDL1 na:mA1 maʔDL1 mitDS2 haʔDL2 koA1 noB1 xawC1 ca:C1 sa:nA1 ja:ŋA1 kujC1 ʔojC1 teŋA1 Bao Yen ka:pDL1 na:mA1 ma:kDL1 pʰakDS1 mjetDS2 tɔ:A1 ra:kDL2 kɔ:A1 nɔ:B1 Cao Bang Lungchow pɯəkDL1 pɤ:kDL1 Shangsi na:mA1 ma:kDL1 pʰakDS1 na:mA1 ma:kDL1 na:mA1 ma:kDL1 pʰakDS1 tɔA1 la:kDL2 kɔA1 nɔB1 tɔ:A1 la:kDL2 ko:A1 no:B1 køA1 nøB1 ruəŋA2 kʰɤwC1 ca:C1 tʰa:nA1 pʰa:ŋC2 ja:ŋA1 kuəjC1 ʔɔ:jC1 pʰɛ:ŋA1 ruəŋA2 kʰ.ɤwC1 ca:C1 ɬa:nA1 pʰa:ŋC1 ja:ŋA1 kuəjC1 ɁɔjC1 pʰjejC1 tʰɛŋA1 ɬu:ŋA2 kʰawC1 kja:C1 ɬa:nA1 ja:ŋA1 ku:jC1 ʔo:jC1 pʰajB1 luŋA2 hawC2 ka:C1 pʰyŋA1 jyŋA1 koyC1 ɁoyC1 pʰajB1 tʰeŋA1 Yay ka:pDL1 ma:kDL1 hokDS1 natDS2 ra:kDL2 koA1 ra:ŋA2 hawC2 caC1 θa:nA1 cuəjC1 ʔojC1 tiəŋA1 Saek pla:kDL1 v ka:pDL1 ma:kDL1 vakDS1 mlɛkDS2 ra:kDL1 na:ŋA2 ruəŋA2 ɣawC2 tla:C1 sa:lA1 viəŋC1 ja:ŋA1 ʔɔ:jC1 pʰrajB1 priəŋA1 332 Gloss 181 ashgourd 182 wild olive 183 starfruit 184 plum 185 pomelo 186 acacia 187 banyan 188 vegetable 189 morning glory 190 mustar green 191 yam 192 taro 193 ginger 194 galangal 195 sesame 196 tea 197 raisin 198 rattan 199 reed 200 grass 201 thatch grass PT *wakD *ko:kD *ɢwɯəŋA *manC *bu:kD *ge:A *rajA *prakD *ɓuŋC *ka:tD *manA *prɯəkD *χi:ŋA *xa:B *r.ŋa:A *ɟa:A *ʔitD *C̥ .wa:jA *ʔo:C *ʰɲɯəC *ɣa:A Siamese fakDS2 kɔ:kDL1 fɯəŋA2 kʰɛ:A2 pʰakDS1 buŋC1 ka:tDL1 manA2 pʰɯəkDL1 kʰiŋA1 kʰa:B1 ŋa:A2 cʰa:A2 wa:jA1 Ɂɔ:C1 ja:C1 kʰa:A2 Sapa fakDS2 manC2 pukDL2 hajB2 fakDS1 -i buŋC1 ka:tDL1 pʰɯʔDL1 xiŋA1 xa:B1 ŋa:A2 ʔitDS1 va:jA1 ʔoC1 ɲa:C1 xa:A2 Bao Yen kɔ:kDL1 manC2 pukDS2 -v Cao Bang Lungchow Shangsi vakDS2 fakDS2 fakDS2 manC1 b̤ ukDL2 fə:ŋA2 pukDL2 fyŋA2 manC1 pʰjakDS1 buŋC1 ka:tDL1 manA2 pʰɯəkDL1 kʰiŋA1 kʰa:B1 ŋa:A2 ʔitDS1 wa:jA1 ʔɔ:C1 ɲa:C1 kʰa:A2 rɤjA2 pʰjakDS1 buŋC1 ka:tDL1 pʰɯəkDL1 kʰa:B1 ŋa:A2 za:A2 ɁitDS1 wa:jA1 ɁɔC1 ɲa:C1 g̈ a:A2 pʰjakDS1 buŋC1 ka:tDL1 manA2 pʰɤ:kDL1 kʰiŋA1 ŋa:A2 ca:A2 va:jA1 Ɂo:C1 ja:C1 ka:A2 boŋC1 ka:tDL1 manA2 pʰykDL1 hɤŋA1 ha:B1 ŋa:A2 sa:A2 jitDS1 wa:jA1 ɁøC1 ɲyC1 ha:A2 Yay kokDL1 pukDS2 keA2 piəkDS1 ka:tDL1 manA2 pɯəkDL1 hiŋA1 haB1 raA2 saA2 va:jA1 ɲiəC1 haA2 Saek vakDS2 kɔ:kDL1 viəŋA1 hajA2 pʰrakDS1 manA2 hi:ŋA1 ha:B1 sa:A2 ɁitDS1 va:jA1 ɲuəC1 ɣa:A2 333 Gloss 202 mushroom 203 fern 204 duckweed 205 aquatic moss D. Nature 206 water 207 fire 208 flame 209 smoke 210 soot 211 hard firewood 212 bamboo firewood 213 leaf ashes 214 ashes (wood) 215 iron (1) 216 iron (2) 217 salt 218 lye 219 grease, fat 220 soil PT *ʰrwetD *ku:tD *ʰne:A *dawA Siamese hetDS1 ku:tDL1 nɛ:A1 tʰawA2 Sapa hitDS1 kutDS1 neA1 tawA2 Bao Yen hɤtDS1 ku:tDL1 nɛ:A1 tɤwA2 Cao Bang kutDL1 nɛA1 d̤ ɤwA2 Lungchow vitDS2 kutDL1 ne:A1 Shangsi litDS1 kutDL1 Yay ratDS1 kutDS1 tawA2 Saek ku:tDL1 nɛ:A1 *C̬ .namC *wɤjA *ple:wA *ɣwanA *ʰmi:C *wɯ:lA *ʰlɯəwA *brawB *dawB *ʰlekD *mwa:A *klwɯəA *ɗaŋB *manA *tɯmA na:mC2 fajA2 ple:wA2 kʰwanA2 mi:C1 fɯ:nA2 tʰawB2 lekDS1 klɯəA1 da:ŋB1 -v manA2 tomA1 namC2 pewA1 vanA2 miC1 funA2 tawB2 likDS1 kɯA1 namC2 pʰɤjA2 pje:wA1 wanA2 mi:C1 pʰɯ:nA2 luəA1 pjɤwB2 tɤwB2 lekDS1 namC2 vɤjA2 pjewA1 wanA2 mi:C1 wɯnA2 luəA1 d̤ oɤwB2 lekDS1 namC2 fajA2 vanA2 mi:C1 fɤnA2 pjawB2 likDS1 namC2 foyA2 pewA1 wanA2 liwA1 tawB2 likDS1 kɯəA1 kɯəA1 daŋB1 manA2 tomA1 kɯ:A1 daŋB1 tumA1 kyA1 daŋB1 tomA1 ramC2 fiA2 miC1 funA2 tawB2 faA2 kuəA1 daŋB1 namC2 vi:A2 ple:wA1 ɣɔnA2 mi:C1 vɯlA1 tʰawB2 ma:A2 tluəA1 daŋB1 manA2 334 Gloss 221 earth 222 sand 223 stone 224 dry land 225 wild 226 forest 227 grove, wood 228 place, ground 229 mountain 230 stone mountain 231 cave 232 hole 233 crack, hole 234 pit 235 well, spring 236 mountain stream 237 creek 238 river 239 river bank 240 beach, sandbar 241 mud PT *ɗin *zwɯəjA *tri:lA *ɓokD *C̬ .tɯənB *ɗoŋA *pa:B *di:B *C̥ .do:jA *pra:A *cramC *ru:A *ɟo:ŋB *C̬ .kumA *ɓo:B *qrwɤjC *ro:ŋB *da:B *ʰwaŋB *ha:tD *lɤmB Siamese dinA2 sa:jA2 hinA1 bokDS1 tʰɯənB1 doŋA1 pa:B1 tʰi:B2 dɔ:jA2 pʰa:A1 tʰamC1 ru:A2 cʰɔŋB2 kʰumA1 bɔ:B1 huəjC1 rɔ:ŋB2 tʰa:B2 faŋB1 ha:tDL1 lomB1 Sapa dinA1 sa:jA2 hinA1 bukDS1 tʰɯənB1 duŋA1 pa:B1 tiB2 pʰa:A1 tʰamC1 huA2 coŋA2 -t boB1 hujC2 -t hoŋB2 ta:B2 Bao Yen dinA1 tʰa:jA2 hi:nA1 bokDS1 tʰuənB1 -v doŋA1 pa:B1 ti:C2 pʰja:A1 tʰamC1 ru:A2 Cao Bang dinA1 ra:jA2 tʰinA1 bokDS1 tʰɯənC1 -t doŋA1 pa:B1 d̤ iB2 dɔjA1 pʰja:A1 tʰamC1 ruA2 Lungchow ɬa:jA2 hinA1 buk5 tʰɤ:nA1 duŋA1 ti:B2 pʰja:A1 ɬu:A2 Shangsi ɬuyA2 tʰɤnA1 tunB2 doŋA1 pa:B1 toyB2 døyA1 pʰa:A1 sumA1 -i bɔ:B1 kʰuəjC1 rɔ:ŋC2 -t pʰaŋB1 ha:tDL1 bɔB1 kʰuəjC1 rɔŋB2 d̤ a:B2 ha:tDL1 lumB1 kʰumA1 bo:B1 vu:jC1 ta:B2 ha:tDL1 lumB1 kʰomA2 bøB1 loŋB1 ta:B2 pʰaŋB1 Yay θiəjA2 rinA1 bokDS1 tɯənB2 doŋA1 tiB2 dojA1 soŋB2 kumA2 boB1 viC1 taB2 lamB1 Saek jo:jA2 ri:lA1 bɔkDS1 tʰuəlB2 dɔŋA1 pa:B1 tʰi:B1 rɔ:jA1 pʰra:A1 tʰamC1 sɔ:ŋB2 kʰumA2 bɔ:B1 ri:C1 tʰa:B2 vaŋB1 ha:tDL2 335 Gloss 242 moon, month 243 star (general) 244 star (in the sky) 245 moonlight 246 sunshine 247 cloud 248 fog 249 wind 250 sky, heaven 251 sky, weather 252 rain 253 gust (of rain) 254 lightning 255 thunder 256 hoarfrost 257 dew, mist 258 hail 259 steam, vapor 260 shade 261 shadow, reflection 262 dry season PT *ɓlɯənA *t.na:wA *ɗi:B *ʰŋa:jA *C̥ .dwi:tD *ʰwɯəC *ʰmo:kD *C̬ .lɯmA *ɓɯnA *va:C *C̥ .wɯnA *kra:B *m.le:pD *pra:C *ʰmɯǝjA *C̬ .nwa:jA *trepD *s.ʔwɤ:jA *rɤmB *ŋawA *C̬ .le:ŋC Siamese dɯənA1 da:wA1 ŋa:jA1 dɛ:tDL1 fa:C1 mɔ:kDL1 lomA2 bonA1 fa:C2 fonA1 ha:B1 lɛ:pDL2 hepDS1 ʔajA2 -v romB2 ŋawA2 lɛ:ŋC2 Sapa bɯənA1 da:wA1 diB1 detDL1 moʔDL1 fa:C2 mujB2 hipDS1 ʔa:jA1 homB2 homB2 leŋC2 Bao Yen bɯənA1 da:wA1 ha:jA1 dɛ:tDL1 mɔ:kDL1 lomA2 bonA1 pʰɤnA1 Cao Bang bɯənA1 da:wA1 diB1 ha:jA1 dɛtDL2 pʰa:C1 mɔkDL1 lomA2 bɤnA1 fa:C2 pʰɤnA1 sa:B1 mjepA2 Lungchow bɤ:nA1 da:wA1 di:B1 ha:jA1 de:tDL1 pʰa:C1 mo:kDL1 lumA2 fa:C2 pʰɤnA1 ha:B1 me:pDL2 Shangsi bunA1 da:wA1 doyB1 ha:jA1 ditDS1 pʰyC1 mokDL1 lomA2 bonA1 pʰonA1 la:B1 mɯəjA1 hopDS1 ʔa:jA1 romB2 ŋɤwA2 lɛ:ŋC2 mɯəjA1 tʰepDS1 ja:jA1 ŋɤwA2 lɛŋC2 na:jA2 hatDS1 -f ja:jA1 ŋawA2 le:ŋC2 muyA1 na:jA2 tʰitDS1 ŋawA2 leŋC2 Yay dɯənA1 da:wA1 diB1 ditDS1 vɯəC1 mokDL1 rumA2 bunA1 hunA1 raB1 pjaC1 ripDS1 θɯəjA1 ramB2 ŋawA2 reŋC2 Saek bliənA1 tra:wA1 ri:tDL1 viəC1 mɔ:kDL1 lumA2 bɯnA1 vɯnA1 ra:B1 pʰra:C1 so:jA1 ŋawA2 Gloss PT Siamese Sapa Bao Yen Cao Bang Lungchow Shangsi Yay Saek 336 E. Society 263 father, man 264 mother; woman 265 older sibling 266 younger sibling 267 paternal grandfather 268 maternal grandfater 269 maternal grandmother 270 child (offspring) 271 great-grandparent 272 nephew, niece, or grandchild 273 great-grandchild 274 parent's older brother 275 parent's older sister 276 father's younger sister 277 mother's younger sibling 278 father's younger brother 279 wife of man's younger brother 280 wife *bo:B *me:B *bi:B *nwo:ŋC *pɯwB *ta:A *na:jA *lɯ:kD *ɟɯəC *ʰla:nA *ʰlenC *luŋA *pa:C *ʔa:A *na:C *ʔa:wA *lɯǝwA *miəA pʰɔ:B mɛ:B2 pʰi:B2 nɔ:ŋC2 pu:B1 ta:A1 ja:jA2 -i lu:kDL2 cʰɯəC2 la:nA1 le:nA1 t luŋA2 pa:C1 ʔa:A1 na:C2 miəA2 poB meB2 piB2 noŋC2 puB1 ta:A1 na:jA2 luʔDL2 cɯC2 la:nA1 linA1 -t pa:C1 ʔa:A1 na:C2 ʔa:wA1 miA2 pɔ:B2 mɛ:B2 pi:C2 -t nɔ:ŋC2 pu:B1 ta:A1 na:jA2 lukDL2 cɯəC2 la:nA1 lɤnA1 -t luŋA2 pa:C1 ʔa:A1 na:C2 ʔa:wA1 luəA2 miəA2 b̤ ɔB2 mɛB2 b̤ iB2 nɔŋC2 puB1 ta:A1 ta:jB1 -i lukDL2 zɯəC2 la:nA1 lɤnC1 luŋA2 pa:C1 Ɂa:A1 na:B1 -t Ɂa:wA1 luəA2 po:B2 me:B2 pi:B2 no:ŋC2 ta:A1 ta:jA1 lukDL2 la:nA1 luŋA2 pa:C1 ʔa:A1 na:B2 t ʔa:wB1 lu:A2 poC2 -t meB2 nøŋC2 ta:A1 ta:jA1 lakDL1 la:nA lɤnC1 pa:C1 na:B2 -t liwA2 poB2 meB2 piC2 -t nuəŋC2 pʰɔ:B2 me:B2 pʰi:B2 nuəŋC2 taA1 ta:jB1 -i lɯkDS2 ta:A1 ʔa:jA2 -i lɯkDL1 la:nA1 la:nA1 luŋA2 paC1 pa:C1 naC2 ʔa:wA1 na:C2 ʔa:wB2 luəwA2 337 Gloss 281 son-in-law 282 daughter-in-law 283 person, human being 284 child (young person) 285 man, male 286 unmarried man 287 unmarried woman 288 girl 289 lady 290 ralated by marriage 291 widowed 292 orphan 293 name 294 master, owner 295 slave (1) 296 slave (2) 297 Vietnamese 298 shaman 299 spirit (1) 300 spirit (2) 301 soul; whorl in the hair PT *C̬ .kɯəjA *baɰC *ɢwɯnA *ɗekD *ʑa:jA *ɓa:wB *sa:wA *ɓɯ:kD *na:ŋA *t.no:ŋA *ʰma:jC *gm̩ .ra:C *ɟɤ:B *ɕɤwC *χɔ:jB *kra:C *ke:wA *ʰmo:A *pri:A *mwa:ŋA *qwanA Siamese kʰɤ:jA1 pʰajC2 kʰonA2 dekDS1 cʰa:j4 ba:wB1 sa:wA1 na:ŋA2 dɔ:ŋA1 ma:jC1 pʰra:C2 cʰɯ:B2 ca:wC1 kʰɔ:jC1 -t kʰa:C1 kɛ:wA1 mɔ:A1 pʰi:A1 kʰwanA1 Sapa paɰC2 kɯnA2 diŋA1 -f ca:jA2 ba:wB1 sa:wA1 na:ŋA2 doŋA1 ma:jB2 -t pa:C2 xa:C1 kewA1 moA1 pʰiA1 xwanA1 Bao Yen kʰɯəjA1 pɤɰC2 kɤnA2 diŋA1 -f ca:jA2 ba:wB1 tʰa:wA1 Cao Bang kʰɯəjA1 b̤ ɤɰC2 g̈ ɤnA2 dekDS1 za:jA2 ba:wB1 ɬa:wA1 Lungchow kʰɤ:jA1 kɤnA2 dikDS1 ca:jA2 ba:wB1 ɬa:wA1 Shangsi honA2 dikDS1 ba:wB1 ɬa:wA1 na:ŋA2 dɔ:ŋA1 ma:jC1 pja:C2 na:ŋA2 dɔŋA1 ma:jC1 b̤ ja:C2 ma:jC1 pja:C2 ma:jC1 cɤwC1 kʰɔ:jC1 -t sa:C1 kɛ:wA1 mɔ:A1 pʰi:A1 kʰwanA1 cɤwC1 kʰɔjC1 -t kɛwA1 mɔA1 pʰi:A1 ma:ŋA2 kʰwanA1 cawC1 kʰo:jC1 -t ke:wA1 mo:A1 pʰi:A1 kʰwanA1 hoyB1 kewA1 ma:ŋA2 wanA1 Yay kɯəjA2 paɰC2 hunA2 θa:jA1 ba:wB1 θa:wA1 bɯkDS1 doŋA1 ma:jB1 soB2 θuC1 hɔjB1 cewA1 moA1 fa:ŋA2 vanA1 Saek kʰo:jA2 kʰwɤ:C2 -i hunA2 sa:jA1 ba:wB1 sa:wA1 bɯkDS1 na:ŋA2 trɔ:ŋA1 ma:jC2 tʰra:C2 sɔ:B2 su:C1 hɔ:jB1 ke:wA1 mɔ:A1 ma:ŋA2 hɔnA1 Gloss 302 village 303 township 304 debt PT *ɓa:nC *mɯəŋA *ʰni:C Siamese ba:nC1 mɯəŋA2 ni:C1 Sapa ba:nC1 mɯəŋA2 Bao Yen ba:nC1 mɯəŋA2 ni:C1 Cao Bang ba:nC1 mɯəŋA2 ni:C1 Lungchow ba:nC1 Shangsi ba:nC1 Yay ba:nC2 niC1 Saek ba:nC1 338 F. Material culture 305 liquor 306 medicine 307 vinegar 308 flour 309 house 310 granary 311 stake 312 eaves 313 door 314 stairs, ladder 315 pillar 316 partition, lid 317 split bamboo flooring 318 window 319 stool 320 tripod *ʰlawC *ˀjɯəA *ʰmi:B *ɓɯəA *rɤ:nA *ˀjɯǝwC *ʰlakD *ʑa:jA *tu:A *ɗrwajA *sawA *hwa:A *wa:kD *ta:ŋB *taŋB *giǝŋA lawC1 ja:A1 rɯənA2 ja:wC1 lakDS1 cʰa:jA2 tu:A1 dajA2 sawA1 fa:A1 fa:kDL2 ta:ŋB1 taŋB1 kʰiǝŋA2 lawC1 ja:A1 miB1 bɯA1 hɯənA2 tuA1 dajA1 sawA1 fa:A1 ta:ŋB1 taŋB1 lɤwC1 ja:A1 mi:B1 bɯəA1 rɯənA2 jawB1 lakDS1 ca:jA2 tu:A1 dwɤjA1 tʰɤwA1 pʰa:A1 pʰa:kDL2 ta:ŋB1 taŋB1 kiəŋA2 lɤwC1 ja:A1 miB1 bɯəA1 rɯənA2 ja:wC1 lakDS1 tuA1 dwajA1 ɬɤwA1 pʰa:A1 va:kDL2 ta:ŋC1 taŋB1 g̈ iəŋA2 lawC1 ja:A1 mi:B1 bɯ:A1 ɬɤ:nA2 tu:A1 dajA1 ɬawA1 pʰa:A1 ta:ŋB1 taŋB1 ki:ŋA2 lawC1 jyA1 moyB1 lunA2 tawA1 dajA1 ɬawA1 lawC1 jiəA1 ra:nA2 jiəwC2 lakDS1 θa:jA2 tuA1 lajA1 θawA1 vaA1 taŋB1 kʰiŋA2 taŋB1 ciəŋA2 lawC1 (ja:A1) bɯəA1 ra:nA2 jiəwC1 lakDS1 ja:jA2 tu:A1 rajA1 va:A1 va:kDL2 taŋB1 kʰiəŋA2 339 Gloss 321 board 322 thing 323 knife 324 machete, big knife 325 chopping board 326 spear 327 crossbow 328 axe 329 chisel 330 hook 331 pliers, thongs 332 handle (of a knife) 333 handle, rod 334 carrying pole (1) 335 carrying pole (2) 336 rope, cord 337 top for spinning 338 basin 339 jug 340 pot 341 bamboo tube PT *pe:nC *χɔ:ŋA *mitD *ɟm̩ .ra:C *χiəŋA *kro:kD *ʰnwɯəC *xwa:nA *siəwB *xɔ:A *gi:mA *ɗa:mC *galA *tra:pD *ɢa:nA *ɟɤ:k *k.ra:ŋB *ʔa:ŋB *krajA *ʰmo:C *baŋ B/C Siamese pɛnC2 kʰɔ:ŋA1 mi:tDL2 -v pʰra:C2 kʰiəŋA1 hɔ:kDL1 na:C1 kʰwa:nA1 siwB1 kʰɔ:A1 kʰi:mA2 da:mC1 kʰanA2 ha:pDL1 kʰa:nA2 cʰɯəkDL2 kʰa:ŋB1 ʔa:ŋB1 hajA1 mɔ:C1 Sapa penC1 xoŋA1 mitDS2 pa:C2 xiəŋA1 xwa:nA1 siwB1 xoA1 kimA2 damB1 ha:pDL1 ka:nA2 cɯʔC2 ʔa:ŋB1 hajA1 Bao Yen mitDS2 pja:C2 Cao Bang Lungchow pe:nC1 Shangsi penC1 høŋA1 b̤ ja:C2 pja:C2 kʰi:ŋA1 hiŋA1 na:C1 kʰwa:nA1 tʰi:wB1 kʰɔ:A1 ki:mA2 da:mC1 na:C1 kʰwa:nA1 ɬiwB1 g̈ imA2 da:mC1 ɬi:wB1 kimA2 da:mC1 ha:pDL1 ka:nA2 kʰa:ŋB1 -i ʔa:ŋB1 kʰajA1 mɔ:C1 baŋC1 tʰa:pDL1 zɯəkDL2 sa:ŋB1 mɔC1 baŋC1 ha:pDL1 ka:nA2 cɤ:kA2 ha:ŋB1 ʔa:ŋB1 mo:C1 ɬiw5 kʰanA1 tʰa:pDL1 ha:nA2 sa:kDL2 la:ŋB1 møC1 Yay mitDS2 saC2 heŋA1 rokDL1 nɯəC1 va:nA1 θiəwB1 cimA2 da:mC2 kanA2 ra:pDL1 ha:nA2 sa:kB2 ca:ŋB1 rajA1 moC1 baŋB1 Saek pe:nC1 ho:ŋA1 tʰa:C2 nuəC1 va:nA1 si:wB1 hɔ:A1 kʰi:mA2 da:mC1 kʰalA2 ra:pDL1 ɣa:nA2 sa:kB2 ʔa:ŋB1 hajA1 mɔ:C1 baŋC2 340 Gloss 342 bag 343 walking cane 344 comb 345 writing, book 346 broom 347 thread (1) 348 thread (2) 349 hemp 350 loom 351 shuttle of loom 352 spool 353 indigo (1) 354 indigo (2) 355 bamboo strip for tying or weaving 356 needle 357 strand (of rope) 358 lacquer 359 bamboo hat 360 puppet, marionette 361 marking, patterns 362 paddy field PT *croŋA *dawC *ʰrwɯ:jA *sɯ:A *ɲu:A *C̥ .da:jC *ʰmajA *pa:nB *trukD *p.rawB *ʰlwu:tD *g.ra:mA *kromC Siamese tʰuŋA1 tʰa:wC2 wi:A1 sɯ:A1 da:jC1 majA1 pa:nB1 hu:kDL1 lɔ:tDL1 kʰra:mA2 hɔmC1 *b.twu:kD tɔ:kDL1 *qemA *kliəwA *rakD *klɤpD *hunB *C̬ .la:jA *na:A kʰemA1 kliǝwA1 rakDS2 hunB1 la:jA2 na:A2 Sapa tʰuŋA1 tawC2 viA1 sɯA1 ɲuA2 majA1 pa:nB1 huʔDL1 ca:mA2 homC1 toʔDL1 ximA1 hakDS2 la:jA2 na:A2 Bao Yen tʰoŋA1 tɤwC2 wi:A1 Cao Bang tʰoŋA1 d̤ ɤwC2 wiA1 ɬɯA1 ɲuA2 Lungchow tʰuŋA1 vi:A1 ɬɯ:A1 Shangsi tawC6 loyA1 ɬoyA1 mɤjA1 pa:nB1 hu:kDL1 pʰjɤwB1 lɔ:tDL1 ca:mA2 sɔ:mC1 pʰjɔ:kDL1 kʰomA1 mɤjA1 pa:nB1 tʰukDL1 tʰɤwB1 lɔtDL1 za:mA2 sɔmC1 tʰɔkDL1 kʰemA1 majA1 pa:nB1 hukDS1 majA1 tʰokDL1 lo:tDL1 kja:mA2 pʰjo:kDL1 kʰimA1 tʰukDL1 hɤmA1 rakDS2 cupDS1 rakDS2 cupDS1 kipDS1 na:A2 la:jA2 na:A2 la:jA2 na:A2 na:A2 Yay rojA1 θɯA1 ŋuA2 da:jC2 majA1 rokDL1 tawB1 sa:mA2 -i romC1 tukDS2 cimA1 cewA1 capDS1 hunB1 ra:jA2 naA2 Saek tʰɔŋA1 hɔ:jA1 sɯ:A1 ra:jC1 hu:kDS2 lu:tDL1 prukDS2 kimA1 tlɛ:wA1 hunB1 la:jA2 na:A2 341 Gloss 363 dry field 364 open field 365 garden 366 ditch 367 dike between rice fields 368 dam 369 water pipe 370 plough 371 yoke 372 part of yoke under neck 373 mortar (1) 374 mortar (2) 375 pestle 376 winnowing basket 377 bran 378 straw, stubble 379 fish hook 380 fish net 381 fishtrap 382 snare 383 gutter, trough PT *rɤjB *doŋB *swɯ:nA *ʰmɯəŋA *ɣalA *hwa:jA *linA *crwajA *ʔe:kD *ʔo:ŋC *grokD *ʰrumA *sa:kD *ɗoŋC *ramA *wɯǝŋA *ɓetD *kre:A *zajA *re:wC *rwɯəŋA Siamese rajB2 tʰuŋB2 suənA1 mɯəŋA1 kʰanA2 fa:jA1 tʰajA1 ʔɛ:kDL1 kʰrokDS2 sa:kDL1 doŋC1 ramA2 fa:ŋA2 betDS1 hɛ:A1 sajA2 rɛ:wC2 ra:ŋA2 Sapa hajB2 tuŋB2 suənA1 fa:jA1 linA2 tʰajA1 ʔeʔDL1 ʔoŋB2 -t saʔDL1 doŋC1 -v hamA2 fɯəŋA2 -v bitDS1 heA1 sajA2 hewC2 ha:ŋA2 Bao Yen rɤjB2 toŋB2 tʰuənA1 mɯəŋA1 kʰanA2 pha:jA1 linB1 -t tʰɤjA1 ʔɛ:kDL1 ʔoŋC1 cokDS2 tʰa:kDL1 doŋC1 ramA2 pʰa:ŋA2 bɤtDS1 kʰɛ:A1 tʰɤjA2 rɛ:wC2 ra:ŋA2 Cao Bang rɤjB2 d̤ oŋB2 ɬuənA1 mɯəŋA1 g̈ anA2 pʰa:jA1 linA2 tʰɤjA1 ɁekDL2 ɁoŋC1 ɬa:kDL2 doŋC1 ramA2 va:ŋA2 betDS1 kʰɛ1 rɤjA2 rɛwC2 ra:ŋA2 Lungchow ɬajB2 ɬu:nA1 mɤ:ŋA1 kanA2 pʰa:jA1 tʰajA1 ʔe:kDL1 kjukDS2 ɬa:kDL1 duŋC1 ɬamA2 fa:ŋA2 bit5 he:A1 ɬajA2 ɬa:ŋA2 Shangsi loyB2 ɬunA1 myŋA1 hanA2 pʰa:jA1 tʰajA1 ɁekDL1 lomA1 doŋC1 lamA2 fyŋA2 leA1 ɬajA2 Yay riB2 toŋB2 θɯənA1 hanA2 va:jA1 linA2 sajA1 ʔekDL1 ʔoŋC2 cokDS2 rumB1 θa:kDL1 ramA2 fɯəŋA2 reA1 θajA2 ruəŋA2 Saek ri:B2 tʰɔŋB2 suənA1 mɯəŋA1 ɣalA2 linA2 tʰajA1 ʔɛ:kDL1 ʔɔŋC1 sa:kDL1 ramA2 viəŋA1 re:A1 jajA2 re:wC2 Gloss 384 cooked in bamboo tube 385 boat 386 raft 387 road 388 track 389 saddle 390 drum PT *ʰla:mA *C̬ .rwɯəA *be:A *ʰrwɤnA *rwu:jA *ʔa:nA *klo:ŋA Siamese la:mA1 rɯəA2 pʰɛ:A2 honA1 rɔ:jA2 ʔa:nA1 klɔ:ŋA1 Sapa la:mA1 hɯA2 peA2 hojA2 ʔa:nA kuŋB1 -t Bao Yen la:mA1 lɯəA2 -i pɛ:A2 Cao Bang la:mA1 lɯəA2 b̤ ɛA2 Lungchow la:mA1 lɯ:A2 Shangsi luA2 peA2 rɔjA2 Ɂa:nA1 cɔŋA1 kjo:ŋA1 loyA2 Ɂa:nA1 køŋA1 Yay ruəA2 peA2 ranA1 riA2 ʔa:nA1 coŋA1 Saek ruəA1 pʰe:A2 ri:A2 ʔa:nA1 tlɔ:ŋA1 342 G. Adjectives 391 red 392 black 393 white 394 green 395 yellow 396 dark (red) 397 gray 398 white-spotted 399 clear, clean 400 dark 401 bright, light (n.) 402 thin (not fat) *C̥ .dwi:ŋA *C̥ .damA *xa:wA *xiəwA *ʰlɯəŋA *klamB *ʰmo:ŋA *ɓla:ŋB *saɰA *mɯ:tD *roŋB *pro:mA dɛ:ŋA1 damA2 kʰa:wA1 kʰiəwA1 lɯəŋA1 klamB1 mɔ:ŋA1 da:ŋB1 sajA1 mɯ:tDL2 ruŋB2 pʰɔ:mA1 deŋA1 damA1 xa:wA1 xewA1 ba:ŋB1 saɰA1 mutDS2 huŋB2 dɛ:ŋA1 damA1 kʰa:wA1 kʰɛ:wA1 lɯəŋA1 bja:ŋB1 tʰɤɰA1 mɯ:tDL2 ruŋB2 dɛŋA1 damA1 kʰa:wA1 kʰɛwA1 kamB1 -v de:ŋA1 damA1 kʰa:wA1 kʰe:wA1 lɤ:ŋA1 da:ŋB1 ɬɤɰA1 ɬaɰA1 ruŋB2 ɬuŋB2 pʰjo:mA1 dɤŋA1 namA1 hewA1 lyŋA1 møŋA1 ɬoyA1 løŋB1 pʰømA1 diŋA1 ha:wA1 hewA1 camB1 da:ŋB1 θaɰA1 roŋB2 pjomA1 ri:ŋA1 ramA1 ha:wA1 he:wA1 rɔ:ŋB2 -i pʰrɔ:mA1 343 Gloss 403 fat 404 long 405 short (not long) 406 big (1) 407 big (2) 408 small 409 low, short (not tall) 410 heavy 411 light (in weight) 412 thick 413 thin (not thick) 414 new 415 old (of living beings) 416 old (of things) 417 senior 418 young, soft 419 raw, not ripe 420 hard 421 coarse, tough 422 deep 423 loose PT *bwi:A *rɯjA *tinB *ʰluəŋA *ɓɯ:kD *no:jC *tamB *ʰnakD *C̥ .bawA *ʰna:A *C̥ .ba:ŋA *ʰmɤ:lB *ke:B *kawB *crawC *ʔwu:nB *C̥ .dipD *k.re:ŋA *ʰɲa:pD *lɤkD *ʰlo:mA Siamese pʰi:A2 ri:A2 luəŋA1 Sapa piA2 hiA2 tinC1 luəŋA1 Bao Yen pi:A2 ri:A2 tɤnC1 luəŋA1 Cao Bang b̤ iA2 riA2 tɤnC1 luəŋA1 Lungchow pi:A2 ɬi:A2 tinC1 lu:ŋA1 Shangsi poyA2 loyA2 tɤnC1 -t nɔ:jC2 tamB1 nakDS1 bawA1 na:A1 ba:ŋA1 majB1 kɛ:B1 kawB1 tʰawC1 ʔɔ:nB1 dipDS1 kʰɛŋA1 ja:pDL1 lɯkDS2 luəmA1 -v nojC2 tamB1 nakDS1 bawA1 na:A1 ba:ŋA1 maɰB1 keB1 kawB1 tʰawC1 ʔonB1 dipDS1 xeŋA1 ɲa:pDL1 lɯkDS2 luəmA1 -v nɔ:jC2 tamB1 nakDS1 bɤwA1 na:A1 ba:ŋA1 mɤɰB1 kɛ:B1 ʔɔ:nB1 dipDS1 kʰɛ:ŋA1 ɲa:pDL1 lɤkDS2 tamB1 nakDS1 bɤwA1 na:A1 ba:ŋA1 mɤɰB1 kɛB1 kawB1 -v tʰɤwC1 ɁonB1 dipDS1 kʰɛŋA1 ɲa:pDL1 lɤkDS2 no:jC1 -t tamB1 nakDS1 bawA1 na:A1 ba:ŋA1 maɰB1 ke:B1 kawB1 tʰawC1 ʔo:nB1 dipDS1 kʰe:ŋA1 ja:pDL1 dakDS1 -i nøyC2 tamB1 nakDS1 bawA1 na:A1 ba:ŋA1 møB1 kawB1 ɁønB1 dipDS1 leŋA1 lakDS2 Yay piA2 rajA2 tinC1 bɯkDS1 tamB1 nakDS1 bawA1 naA1 ba:ŋA1 moB1 ceB1 kawB1 ʔunB1 dipDS1 ɲa:pDL1 lakDS2 lomA1 Saek pʰu:jA2 rajA2 tinB1 luəŋA1 bɯkDS1 tamB1 nakDS1 vawA1 na:A1 va:ŋA1 mɔ:B1 ke:B1 kawB1 tʰawC1 ʔu:nB1 ripDS1 tʰriəŋA1 ɲiəpDL1 -v lakDS2 344 Gloss 424 tight, narrow 425 steep 426 stuck 427 straight 428 crooked 429 hot 430 warm 431 blind 432 night blind 433 deaf 434 bitter 435 astringent in taste 436 sour 437 sweet, delicious 438 insipid 439 core 440 fragrant 441 stinky 442 putrid 443 fishy 444 empty PT *gapD *ʰliŋB *ga:A *zɤ:B *gotD2 *rwu:lC *ʔunB *ɓo:tD *ʰwa:ŋA *ʰnuəkD *C̬ .qɤmA *ʰwɯətD *sɤmC *C̥ .wa:lA *cɯ:tD *ke:lB *ho:mA *ʰmenA *xwi:wA *ɣa:wA *plɤwB Siamese kʰapDS2 kʰa:A2 sɯ:B2 kʰotDS2 rɔ:nC2 ʔunB1 bɔ:tDL1 fa:ŋA2 nuəkDL1 kʰomA1 fa:tDL1 somC1 wa:nA1 cɯ:tDL1 kɛnB1 hɔ:mA1 menA1 kʰiəwA1 kʰa:wA2 pla:wB1 Sapa kapDL2 liŋB1 ka:A2 sɯB2 honC2 botDL1 fa:ŋA2 fa:tDL1 va:nA1 cutDS1 -v kenB1 homA1 minA1 xiwA1 xa:wA2 pawB1 Bao Yen kapDS2 liŋB1 Cao Bang g̈ apDS2 g̈ a:A2 Lungchow kapDS2 liŋB1 ɬɯ:B2 kutDS2 Shangsi kʰapDS2 lɤŋB1 ɬøB2 ʔunB1 bɔ:tDL1 pʰa:ŋA2 nuəkDL1 kʰomA1 pʰa:tDL1 tʰomC1 wa:nA1 cu:tDL1 hɔ:mA1 mɤnA1 kʰwiwA1 kʰa:wA2 pjɤwB1 ɁunB1 bɔtDL1 va:ŋA2 kʰomA1 ɬomC1 wa:nA1 cutDL1 kɛnB1 hɔmA1 menA1 kʰiwA1 pjɤwB1 ʔunB1 bo:tDL1 nu:kDL1 kʰumA1 ɬumC1 va:nA1 ho:mA1 kʰi:wA1 pjawB1 fa:ŋA2 nukDL1 hamA2 fa:tDL1 ɬomC1 wa:nA1 hømA1 hiwA1 Yay liŋB1 kaA2 θoB2 botDL1 nukDS1 -v hamA2 vɯətDL1 θamC1 va:nA1 homA1 ha:wA2 pjuB1 Saek kʰa:A2 jɔ:B2 ru:lC2 nu:kDL1 ɣamA2 viətDL1 samC1 va:lA1 ke:lB1 kʰi:wA1 ɣa:wA2 plu:B1 345 Gloss 445 full 446 deficient 447 good 448 bad 449 long (of time) 450 slow 451 late 452 near 453 far 454 face down 455 face up 456 sharp-pointed 457 sharp, sharp edge 458 slippery, smooth 459 sticky 460 liquid, soft 461 rotten 462 withered 463 dried up 464 dry 465 wet PT *k.temA *bro:ŋB *ɗɤjA *rwɤ:jC *hɤŋA *na:nA *ʰla:C *k.raɰC *k.lajA *qwamC *ʰŋa:jA *ʰle:mA *ɣɤmA *m.lɯ:lB *ʰniəwA *ʰle:wA *nawB *ʰriəwB *ʰre:ŋC *χaɰB *domA Siamese temA1 pʰrɔŋB2 di:A1 -v ra:jC2 hɯŋA1 na:nA2 la:C1 klajC1 klajA1 kʰwamC1 ŋa:jA1 lɛ:mA1 kʰomA2 lɯ:nB2 niəwA1 le:wA1 nawB2 hiəwB1 hɛ:ŋC1 Sapa timA1 diA1 ha:jC2 hɯŋA1 na:nA1 -t la:C1 saɰC1 xamC1 -w lemA1 munB2 hewB1 heŋC1 Bao Yen tɤmA1 pjɔ:ŋB2 dɤjA1 ra:jC2 hɤŋA1 na:nA2 la:C1 sɤɰC1 kwɤjA1 kʰwamC1 ha:jA1 kʰomA2 mjɯ:nB2 niəwA1 nɤwA2 hɛ:wB1 kʰɤɰB1 Cao Bang temA1 b̤ jɔŋB2 dɤjA1 ra:jC2 hɤŋA1 na:nA2 la:C1 sɤɰC1 kwɤjA1 Lungchow timA1 pjo:ŋB2 dajA1 ɬa:jC2 hɤŋA1 la:C1 kʰjaɰC1 kwajA1 -i kʰumC1 Shangsi lɤmA1 -i doyA1 koyC1 kajA1 g̈ omA2 nɤwB2 kumA2 lɤ:nB1 -t li:wA1 he:wB1 niwA1 lewA1 nawB2 hoy B1 -v kʰɤɰB1 kʰaɰB1 tumA2 hoyB1 tomA2 Yay rimA1 pjoŋB2 diA1 rɯəjC2 na:nA2 caɰC1 cajA1 hamC1 ha:jA1 lemA1 niəwA1 nawB2 rewB1 re:ŋC1 Saek rimA1 pʰrɔ:ŋB2 di:A1 ra:jC2 tlɤ:C1 tlajA1 ŋamC1 -i ŋa:jA1 ɣamA2 mlɯ:lB2 niəwA1 lɛ:wA1 nawB2 hɛ:wB1 -i hɛ:ŋC1 -i v hɯ:B1 -v 346 Gloss 466 mute 467 dumb 468 easy 469 drunk (1) 470 drunk (2) 471 sterile 472 entangled 473 blistered 474 expensive 475 familiar 476 asleep 477 hungry 478 satiated 479 swollen 480 stiff and tired 481 tired, exhausted 482 lazy 483 tired, bored 484 idle, free 485 disgusted 486 crazy PT *ŋwamC *ɓɤɰC *ŋa:jB *mawA *mwi:A *ʰmanA *ɲuŋB/C *bo:ŋA *be:ŋA *gunC *ɗakD *ˀjɯǝkD *ʔi:mB *gaɰB *mɯəjB *ʰnɯəjB *kli:kD *ʰna:jB *ɗwɤ:jA *ɓɯəB *ɓa:C Siamese Sapa bajC1 ŋa:jB2 mawA2 ŋa:jB2 mawA2 manA1 juŋB2 pʰɔ:ŋA2 pʰɛ:ŋA2 kʰunC2 manA1 peŋA2 ja:kDL1 ʔimB1 mɯəjB2 nɯəjB1 ʔimB1 kaɰC1 -t mujB2 nujB1 na:jB1 da:jA1 bɯəB1 ba:C1 na:jB1 bɯB1 ba:C1 Bao Yen wamC ŋa:jB2 mɤwA2 manA1 ɲuŋB2 pɔ:ŋA2 Cao Bang bɤɰC1 ŋa:jB2 manA1 Lungchow wamC mawA2 juŋC1 -t Shangsi wamC mawA2 moyA2 b̤ ɛŋA2 pe:ŋA2 peŋA2 dakDS1 ja:kDL1 ʔi:mB1 kɤɰB2 mɯəjB2 nɯəjB1 dakDS1 ja:kDL1 ɁimB1 g̈ ɤɰB2 nɯəjB1 na:jB1 da:jA1 bɯəB1 ba:C1 da:jA1 bɯəB1 ba:C1 dakB1 ja:kDS1 ʔimB1 kaɰB2 na:jB1 da:jA1 bɯ:B1 ba:C1 dakDS1 jykDL1 ɁɤmB1 kʰoyB2 kikDL1 na:jB1 duyA1 byB1 Yay ŋamC2 ŋa:jB2 manA1 ɲuŋC2 poŋA2 peŋA2 kunC2 dakDS1 jiəkDL1 ʔimB1 mɯəjB2 cikDS1 na:jB1 bɯəB1 Saek ŋa:jB2 mawA2 manA1 ɲuŋC2 pʰɔ:ŋA2 kʰunC2 dakDS1 juəkDS2 ʔi:mB1 tli:kDL1 na:jB1 dɤ:jA1 347 Gloss 487 hurt (1) 488 hurt (2) 489 torn 490 plugged 491 bruised (1) 492 bruised (2) 493 lost 494 alive 495 strong, strength H. Verbs 496 weave, to (cloth) 497 dye, to 498 sew, to 499 embroider, to 500 weave, to (baskets, mats) 501 sow, to; scatter, to 502 transplant, to 503 plant, to 504 sift, to 505 imprison, to PT *cepD *ke:tD *qa:tD *sakA *ɟamC *wokD *ʰloŋA *ɲaŋA *re:ŋA Siamese cepDS1 Sapa cipDS1 kʰa:tDL1 xa:tDL1 cʰamC2 fokDS2 loŋA1 jaŋA2 rɛ:ŋA2 luŋA1 heŋA2 Bao Yen cepDS1 kʰa:tDL1 camC2 pʰokDS2 loŋA1 ɲaŋA2 rɛ:ŋA2 Cao Bang cepDS1 kʰa:tDL1 ɬakDS1 Lungchow cipDS1 kʰa:tDL1 Shangsi ketDL1 sakDS1 -i vokDS2 loŋA1 ɲaŋA2 rɛŋA2 luŋB1 ɬe:ŋA2 fokDL2 -v leŋA2 Yay cetDL1 ka:tDL1 camC2 fokDS2 loŋA1 reŋA2 Saek ke:tDL1 ka:tDL1 samC2 vɔkDS2 lɔŋA1 re:ŋA2 *tamB *ɲwu:mC *ɲepD *se:wB tamB1 jɔ:mC2 jepDS2 *sa:n sa:nA1 *C̥ .wa:lB *t.namA *plwɯ:kD *qrɤŋA *k.raŋA wa:nB1 damA1 plu:kDL1 kʰɯŋA1 kʰaŋA1 tamB1 ɲomC2 ɲapDS1 sewB1 sa:nA1 va:nB1 puʔDL1 sɯŋA1 saŋA1 tamB1 ɲɔ:mC2 ɲapDS2 -v tʰɛ:wB1 tʰa:nA1 wa:nB1 damA1 sɤŋA1 saŋA1 tamB1 ɲɔmC2 ɲapDS2 ɬɛwB1 ɬa:nA1 wa:nA1 damA1 pjukDL1 sɤŋA1 saŋA1 tamB1 jo:mC2 japDS2 -v tamB1 ɲomC2 ɲipDS2 ɬa:nA1 va:nB1 damA1 pjɤ:kDL1 kʰjɤŋA1 haŋA1 ɬa:nA1 wa:nB1 damA1 pokDL1 laŋA1 tamC1 ɲumC2 ɲipDS2 θa:nA1 damA1 caŋA1 tamB1 ɲumC2 ɲipDS2 sa:nA1 va:lB1 tramA1 plɔ:kDL1 raŋA1 tʰraŋA1 348 Gloss PT 506 hunt, to *p.rawB 507 mark, to; aim, to *ʰma:jA 508 shoot, to *ɲɯŋA 509 poison (fish), to *C̥ .bɯəA 510 tend animal, to *ɟɯǝŋC 511 raise, to *liəŋC 512 make noise, to *ɗaŋA 513 differ, to *ta:ŋB 514 count, to *napD 515 heap up, to *ko:ŋA 516 weigh, to *ɟaŋB 517 announce, to *pa:wB 518 answer, to *χa:nA 519 ask, to *c.ra:mA 520 beg, to *k.ro:A 521 order, to; blow the nose, to *saŋB 522 scold, to; revile, to *ɗa:B 523 warn,to *tɯǝnA 524 weep, to *t.hajC 525 moan, to *gra:ŋA 526 bark, to *ʰrawB Siamese ma:jA1 jiŋA2 bɯəA1 liəŋC2 daŋA1 ta:ŋB1 napDS2 kɔ:ŋA1 cʰaŋB2 pa:wB1 kʰa:nA1 tʰa:mA1 kʰɔ:A1 saŋB1 da:B1 tɯǝnA1 ha:jC1 kʰra:ŋA2 hawB1 Sapa tawB1 ma:jA1 bɯA1 cɯəŋC2 liəŋC2 pa:wB1 xa:nA1 tʰa:mA1 soA1 saŋB1 da:B1 ca:ŋA2 hawB1 Bao Yen pʰjɤwB1 ma:jA1 bɯəA1 liəŋC2 daŋA1 napDS2 kɔ:ŋA1 pa:wB1 kʰa:nA1 tʰa:mA1 sɔ:A1 tʰaŋB1 Cao Bang ma:jA1 bɯəA1 zɯəŋC2 liəŋC2 daŋA1 ta:ŋB1 Lungchow pʰjawB1 jaŋA2 bɯ:A1 cʰɤ:ŋC2 ta:ŋB1 Shangsi syŋC2 pa:wB1 kʰa:nA1 sa:mA1 sɔA1 ɬaŋB1 da:B1 saŋB2 tʰa:mA1 ho:A1 ɬaŋB2 da:B1 tʰa:mA1 løA1 ɬaŋB1 da:B1 hajC1 ca:ŋA2 hawB1 hajC1 za:ŋA2 hawB1 -v hajC1 hawB1 hajC1 hawB1 Yay tawB1 ma:jA1 ŋiŋA2 siəŋC2 daŋA1 ta:ŋB1 koŋA1 saŋB2 pa:wB1 ha:nA1 sa:mA1 θaŋB1 daB1 tɯənB1 tajC1 rawB1 Saek ma:jA1 ŋiŋA2 viəA1 liəŋC2 ta:ŋB1 napDS2 kɔ:ŋA1 saŋB2 pa:wB1 ŋa:nA1 -i tʰa:mA1 tʰrɔ:A1 saŋB1 hawB1 -i 349 Gloss 527 crow, to 528 laugh, to 529 recite, to 530 open (the mouth), to 531 eat, to 532 chew, to 533 lick, to 534 put out of the mouth, to 535 vomit, to 536 take a bite, to 537 hold in the mouth, to 538 bite, to 539 nibble, to; peck, to 540 kiss, to 541 yawn, to 542 blow, to 543 snore, to 544 cough, to 545 hold the breath, to 546 hug, to; embrace, to 547 build, to PT *χalA *krɯəwA *ʔa:nB *ʔa:C *kɯɲA *giəwC *C̬ .lwiəA *ɣa:jA *rwɯəkD *katD *ʔɤmA *C̬ .qɤpD *tɔ:tD *cu:pD *ʰra:wA *powB *klɤlA *ʔajA *klanC *ko:tD *ko:B Siamese kʰanA1 huəA1 ʔa:nB1 ʔa:C1 kinA1 kʰiəwC2 liəA2 kʰa:jA2 ra:kDL2 katDS1 ʔomA1 kʰopDS1 tɔ:tDL1 cu:pDL1 ha:wA1 pawB1 kronA1 -i ʔajA1 klanC1 kɔ:tDL1 kɔ:B1 Sapa xanA1 ʔa:nB1 ʔa:C1 kinA1 kewC2 xa:jA2 haʔDL2 katDS1 ʔumA1 xupDS1 totDL1 cupDS1 ha:wA1 konA1 ʔajA1 kotDL1 Bao Yen kʰanA1 kʰuəA1 ʔa:nB1 ʔa:C1 kinA1 kɛ:wC2 kʰa:jA2 ra:kDL2 ʔomA1 kʰopDS1 tɔ:tDL1 cu:pDL1 ha:wA1 pɤwB1 cɤnA1 ʔajA1 canC1 kɔ:tDL1 kɔ:B1 Cao Bang kʰanA1 kʰuəA1 Ɂa:nB1 Ɂa:C1 kinA1 hgɛwC2 liA2 -v ga:jA2 ra:kDL2 ɁomA1 kʰopDS1 tɔtDL2 cupDL1 ha:wA1 pɤwB1 canA1 -v ɁajA1 Lungchow kʰanA1 hu:A1 ʔa:nB1 ʔa:C1 kinA1 li:A2 ɬa:kDL2 katDS1 kʰo:pDS1 to:tDL1 cupDL1 ha:wA1 pawB1 ʔajA1 Shangsi hanA1 liwA1 kɤnA1 kʰewC2 liA2 lukDL2 hapDS2 totDL1 cupDL1 ha:wA1 pøB1 kanA1 ɁajA1 kɔtDL1 kɔB1 ko:tDL1 kotDL1 Yay hanA1 riəwA1 ʔaC2 kɯnA1 cewC2 riəA2 ha:jA2 ruəkDL2 katDS1 hapDS2 totDL1 supDS1 ra:wA1 poB1 ʔajA1 canC1 kotDL1 koB1 Saek halA1 ruəwA1 ʔa:nB1 ʔa:C1 kinA1 liəA2 ɣa:jA2 ruəkDL2 katDS1 ʔamA1 ɣapDS2 tɔ:tDL1 cu:pDL1 ra:wA1 pʰɔ:B2 tlɛlA1 ʔajA1 kɔ:tDL1 kɔ:B1 350 Gloss 548 close (the eyes), to 549 lie down, to, sleep, to 550 incubate, to 551 hold in the jaws, to 552 hold in the closed hand, to 553 take up in the two cupped hands, to 554 pick up, to 555 grasp, to; perch, to 556 squeeze, to 557 pinch, to 558 raise, to 559 lift, to 560 carry in the arms, to 561 carry (a bag), to 562 carry on the back, to 563 pull, to; drag, to 564 bend over, to 565 turn upside down or inside out, to 566 lean back, to 567 point, to PT *ʰlapD *nwɯ:nA *wakD *ga:pD *kamA *ko:pD *ˀjipD *capD *galC *ʰni:pD *ɲɤkD *ɲo:A *ʔu:mC *tri:wC *trwa:mA *C̬ .la:kD *kɤmC *pli:nC *ʔi:ŋA *ɟi:C Siamese lapDS1 nɔ:nA2 fakDS2 kʰa:pDL2 kamA1 kɔ:pDL1 jipDS1 capDS1 kʰanC2 ni:pDL1 jokDS2 jɔ:A2 ʔumC1 hiwC1 ha:mA1 la:kDL2 komC1 plinC1 ʔiŋA1 cʰi:C2 Sapa lapDS1 nonA2 ka:pDL2 kamA1 kopDL1 jipDS1 capDS1 ɲoA2 ʔumC1 hiwC1 ha:mA1 laʔDL2 kumC1 ʔiŋA1 ciC2 Bao Yen lapDS1 nɔ:nA2 pʰakDS2 kamA1 Cao Bang lapDS1 nɔnA2 vakDS2 g̈ a:pDL2 Lungchow lapDS1 no:nA2 fakDS2 Shangsi nønA2 fakDS2 kamA1 kamA1 kamA1 Yay lapDS1 ninA2 -v fakDS2 ka:pDL2 kamA1 capDS1 kanC2 kɔpDL1 ɁepDS1 capDS1 ganC2 ɲakDS2 -v ɲɔ:A2 ʔomC1 hiwC1 ha:mA1 la:kDL2 komC1 pi:nC1 ʔiŋA1 ci:C2 jɔA2 -i ɁumC1 tʰiwC1 tʰa:mA1 la:kDL2 komC1 ɁiŋA1 ziC2 ʔipDS1 capDS1 nipDL1 ʔumC1 ha:mA1 la:kDL2 kumC1 kopDL1 jipDS1 nipDL1 ɁomC1 tʰiwC1 tʰa:mA1 komC1 nipDS1 jakDS2 joA1 ʔumC1 riəwC1 ra:mA1 ra:kDL2 pinC1 ʔiŋA1 Saek nu:nA2 kʰa:pDL2 kamA1 capDS1 kʰalC2 ni:pDL1 ɲukDS2 ʔu:mC1 ri:wC1 ra:mA1 kamC1 pli:nC1 ʔiŋA1 -v 351 Gloss 568 stroke, to; carress, to 569 scratch, to 570 put, to 571 snap, to 572 stamp (on), to 573 step on, to (1) 574 step on, to (2) 575 repeat, to 576 tremble, to 577 shake, to 578 wrestle, to 579 twist, to; wring, to 580 stretch out, to 581 sit, to 582 come, to 583 go, to 584 arrive, to 585 go upward, to 586 ascend to 587 descend, to 588 enter, to PT *C̬ .lu:pD *kawA *s.cɤ:lB *ɓli:tD *dɯ:pD *ɲamB *ˀjiǝpD *zamC *salB *C̥ .wajA *plamC *ɓitD *ˀjiətD *naŋB *ʰma:A *pajA *C̬ .tɤŋA *mɯəA *χɯnC *n̩ .loŋA *χawC Siamese lu:pDL2 kawA1 sajB1 di:tDL1 tʰɯ:pDL2 jamB2 jiǝpDS1 samC2 sanB1 wajA1 plamC1 bitDS1 jiətDL1 naŋB2 ma:A2 -i pajA1 tʰɯŋA1 mɯəA2 kʰɯnC1 loŋA2 kʰawC1 Sapa lupDS2 saɰB1 bitDS1 samC2 sanB1 naŋB2 ma:A2 -i pajA1 tʰɯŋA1 xawC1 Bao Yen lu:pDL2 tʰɤɰB1 Cao Bang lupDL2 Lungchow kawA1 Shangsi lupDL2 kawA1 Yay rupDS2 kawA1 soB1 ɲamB2 ɲamB2 jamB2 ɲamB2 vɤjA1 pjamC1 jetDS1 naŋB2 ma:A2 -i pɤjA1 tʰɤŋA1 mɯəA2 kʰɯnC1 loŋA2 kʰawC1 ɬanB1 wɤjA1 bitDS1 jiətDL2 naŋB2 ma:A2 -i pɤjA1 tʰɤŋA1 mɯəA2 kʰɯnC1 loŋA2 kʰɤwC1 bitDS1 ji:tDL1 naŋB2 ma:A2 -i pajA1 tʰɤŋA1 mɯ:A2 kʰɤnC1 nuŋA2 kʰawC1 bitDS1 jitDL1 naŋB1 ma:A2 -i poyA1 noŋA2 hawC1 θanB1 naŋB2 maA1 pajA1 taŋA2 hɯnC1 roŋA2 hawC1 Saek lu:pDL2 kawA1 cɔ:B1 bli:tDL1 tʰɯ:pDL2 jiəpDS2 jamC2 sɛlB1 plamC1 bitDS1 naŋB2 ma:A1 pajA1 tʰaŋA2 mɯəA2 hɯnC1 lɔŋA2 hawC1 352 Gloss 589 exit, to 590 cross, to 591 cross over, to 592 drop, to 593 fall off, to 594 fall down, to 595 slip and fall, to 596 stand, to 597 walk, to 598 fly, to 599 crawl, to 600 crawl over, to 601 climb, to 602 flow, to 603 flood, to 604 spill, to 605 pour, to 606 swim, to 607 float, to 608 sink, to 609 dive, to PT *ʔo:kD *χa:mC *kwa:B *tokD *ʰlɤnB *lɤmC *bla:tD *C̥ .ɟɯ:nA *pra:jC *ɓilA *g.lwɤ:nA *twajB *pi:nA *ʰlwajA *C̬ .tuǝmB *ɓa:B *rwa:tD *lo:jA *wu:A *cɤmA *ɗamA Siamese ʔɔ:kDL1 kʰa:mC1 kwa:B1 tokDS1 lonB1 lomC2 pʰla:tDL2 jɯ:nA1 pʰa:jC1 binA1 kʰla:nA2 ta:jB1 pi:nA1 lajA1 tʰuǝmB2 ba:B1 ra:tDL2 lɔ:jA2 fu:A2 comA1 damA2 Sapa ʔoʔDL1 xa:mC1 kwa:B1 tukDS1 lonC1 -t lumC2 pʰa:tDL2 -i jinA1 fa:jC1 -i binA1 ca:nA2 pinA1 lajA1 tʰumC1 ba:B1 lojA2 cumA1 damA1 Bao Yen ʔɔ:kDL1 kʰa:mC1 kwa:B1 tokDS1 lomC2 pja:tDL2 Cao Bang ɁokDL1 kʰa:mC1 kwa:B1 tokDS1 lɤnB1 lomC2 b̤ ja:tDL2 Lungchow ʔo:kDL1 kʰa:mC1 kwa:B1 tukDS1 Shangsi ɁokDL1 ha:mC1 kwa:B1 tokDS1 lonB1 lomC2 bɤnA1 tɤjB1 pi:nA1 lwɤjA1 tʰuəmC1 ba:B1 lɔ:jA2 pʰu:A2 pʰja:jC1 benA1 -v za:nA2 twɤjB1 pinA1 lwɤjA1 tʰuəmC1 ba:B1 rwa:tDL2 lɔjA2 vu:A2 pʰja:jC1 binA1 kja:nA2 lajA1 tʰu:mC1 ba:B1 fu:A2 pʰa:jC1 bɤnA1 lunA2 lajA1 tomC2 foA2 -v damA1 damA1 Yay ʔokDL1 ha:mC1 kwaB1 tokDS1 lanB1 lamC2 pja:tDL2 dɯnA1 pja:jC1 binA1 rɯənA2 pinA1 lajA1 tumC2 fuA2 samA1 Saek ʔɔ:kDL1 ha:mC1 kwa:B1 tɔkDS1 lamC2 bɯlA1 -v luənA2 tajB1 pi:nA1 tʰumB1 rɔ:tDL1 vu:A1 camA1 353 Gloss 610 close, to 611 revolve, to; spin, to 612 go hungry, to 613 hate, to 614 quit, to 615 see, to 616 hear, to 617 taste, to 618 smell (tr.), to 619 forget, to 620 dream, to 621 believe, to 622 recognize, to 623 know, to 624 be, to; b̩ ecome, to 625 be (in a place), to 626 fear, to (1) 627 fear, to (2) 628 itch, to (1) 629 itch, to (2) 630 seek, to PT *hapD *panB *ʔɤtD *ɟaŋA *ˀja:B *tranA *ŋinA *ɟimA *ɗɤmA *lɯ:mA *hwanA *ɟɯǝB *cakD *rɯ:wC *m̩ .pelA *ˀju:B *ʰla:wA *ja:nC *ɣalA *ɣomA *kraA Siamese hapDS1 panB1 ʔotDS1 cʰaŋA2 ja:B1 henA1 -v jinA2 cʰimA2 domA1 lɯ:mA2 fanA1 cʰɯǝB2 cakDS1 ru:C2 penA2 ju:B1 Sapa hapDS1 panB1 caŋA2 hinA1 -v ŋinA2 cimA2 dumA1 lumA2 cakDS1 huC2 pinA1 juB1 janC1 -v kʰanA2 ja:nC1 xanA2 ha:A1 Bao Yen panB1 ʔotDS1 caŋA2 hanA1 ŋinA2 ci:mA2 domA1 lɯ:mA2 pʰanA1 Cao Bang hapDS1 Lungchow panB1 Shangsi hapDS1 panB1 zaŋA2 ja:B1 hanA1 ɲinA2 zimA2 caŋA2 hanA1 hinC2 cimA2 saŋA2 tʰɤnA1 -v sɤmA2 lumA2 pʰanA1 lumA2 lomA2 pʰanA1 Yay hapDS1 panB1 ʔot4 saŋA2 ranA1 simA2 lumA2 ru:C2 pɤnA1 ju:B1 la:wA1 ja:nC1 kʰanA2 sa:A1 cakDS1 b̤ enA2 juB1 la:wA1 ja:nC2 sa:A1 ɬu:C2 pinA2 ju:B1 la:wA1 ja:nC1 ha:A1 løC2 jawB1 homA2 roC2 panA2 juB1 la:wA1 humA2 raA1 Saek saŋA2 rɛnA1 damA1 vanA2 -t sɯəB2 cakDS1 rɔ:C2 pʰalA2 ju:B1 la:wA1 ɣalA2 ɣumA2 ra:A1 354 Gloss 631 join, to 632 divide, to 633 separate, to 634 castrate, to 635 cut, to 636 slash, to 637 chop, to 638 weed, to 639 break, to 640 burst, to 641 pluck, to 642 take down, to; put down, to 643 peel, to 644 hammer, to 645 pound, to 646 pound (rice), to 647 slap, to 648 pound, to (2) 649 pound, to (1) 650 come into contact, to 651 tie, to PT *to:B *panA *bra:kD *to:nA *tacD *walA *tramC *ɓla:jA *t.rakD *p.re:kD *C̥ .decD *ploŋA *po:kD *ɣo:lC *tamA *zo:mC *tɤpD *dupA *to:kD *C̬ .tɯ:kD *cm̩ .rukD Siamese tɔ:B1 panA1 pʰra:kDL2 tɔ:nA1 tatDS1 fanA2 hamC1 da:jA1 hakDS1 tɛ:kDL1 detDS1 ploŋA1 pɔ:kDL1 kʰɔ:nC2 tamA1 sɔ:mC2 topDS1 tʰup4 tɔ:kDL1 tʰu:kDL1 pʰu:kDL1 Sapa panA1 tonA1 tatDS1 fanA2 hamC1 ba:jAA1 hakDL2 teʔDL1 poʔDL1 tamA1 somC2 tupDS1 toʔDL1 tʰuʔDL1 fuʔDL1 -i Bao Yen tɔ:B1 pja:kDL2 tɔ:nA1 pʰanA2 hamC1 bja:jA1 hakDS1 pʰɛ:kDL1 Cao Bang tɔB1 b̤ ja:kDL2 tɔnA1 tatDS1 vanA2 tʰamC1 bja:jA1 takDS1 tʰeɛkDL1 Lungchow panA1 pja:kDL2 to:nA1 tatDS1 tʰamC1 bja:jA1 takDS1 pʰe:kDL1 Shangsi tatDS1 tʰamA1 -t ba:jA1 tʰakDS1 tʰekDL1 Yay toB1 pja:kDL2 tonA1 tatDS1 fanA2 ramC1 da:jA1 rakDS1 tekDL1 pɔ:kDL1 kʰɔ:nC2 tamA1 tʰɔ:mC2 tɤpDS1 tɔ:kDL1 tʰu:kDL1 pʰju:kDL1 pɔkDL1 tamA1 topDS1 d̤ upDS2 tɔkDL1 tʰukDL1 pʰu:kDL1 pjuŋB1 po:kDL1 ko:nC2 ɬo:mC2 tupDS1 tʰukDL1 pʰukDS1 tamA1 sukDL2 pjoŋA1 pokDL1 honC2 tamA1 tupDS2 tokDL1 tɯkDS2 sukDS2 Saek tɔ:B1 panA1 tɔ:nA1 tɛkDS1 valA1 -t rakDS1 pre:kDL1 rɛkDS1 plɔŋA1 ɣɔ:lC2 tamA1 tapDS1 tɔ:kDL1 tʰɯkDL1 355 Gloss 652 tie up, to 653 undo, to; untie, to 654 fold, to 655 fold double, to 656 stretch, to 657 erect, to 658 plug (a hole), to 659 salt, to 660 pickle, to 661 soak, to (2) 662 scoop, to 663 soak, to (1) 664 burn (intr.), to 665 burn (tr.), to 666 roast, to 667 boil (tr.), to 668 boil (intr.), to 669 scald, to 670 cook (rice), to 671 steam, to 672 make a fire, to PT *la:mB *ke:C *bapD *dopD *ɲɯ:tD *taŋC *ʔwɯtD *ʔɯǝpD *ɗo:ŋA *ʰma:B *takD *ɟe:B *ʰmajC *prawA *ˀjɯəŋC *tomC *ɗɤ:tD *luəkD *truŋA *ʰnɯŋC *ɗaŋA Siamese la:mB2 kɛ:C1 pʰapDS2 tʰopDS2 jɯ:tDL2 taŋC1 ʔutDS1 Sapa keC1 tupDS2 taŋC1 ʔutDS1 dɔ:ŋA1 ma:B1 takDS1 cʰɛ:B2 majC1 pʰawA1 ja:ŋC1 tomC1 dɯǝt2 luəkDL2 huŋA1 nɯŋC1 ma:B1 takDS1 majC1 ja:ŋB2 tumC1 luʔDL2 nuŋA2 -t daŋC1 -t Bao Yen la:mB2 kɛ:C1 topDS2 jɯ:tDL2 -i taŋC1 ʔotDS1 ʔɯəpDL1 Cao Bang la:mB2 kɛC1 b̤ apDS2 d̤ opDS2 ɲutDL2 taŋC1 ɁotDS1 ɁɯəpDL1 Lungchow la:mB2 ke:C1 taŋC1 ʔutDS1 ʔɤ:pDS1 Shangsi la:mB2 topDS2 ma:B1 takDS1 cɛ:B2 mɤjC1 pʰjɤwA1 ja:ŋB2 tomC1 luəkDL2 huŋA1 nɯŋC1 daŋA1 ma:B1 takDS1 zɛB2 mɤjC1 pʰjɤwA1 ja:ŋC1 -t tomC1 dɯətDL1 huŋA1 nɯŋC1 daŋA1 takDS1 majC1 tumC1 dɤ:tDL1 lu:kDL2 huŋA1 daŋA1 takDS1 tomC1 dutDL1 tʰoŋA1 naŋC1 daŋA1 Yay papDS2 taŋC1 doŋA1 maB1 takDS1 seB2 piəwA1 jiəŋC2 tumB1 da:tDL1 lukDS1 ruŋA1 naŋC1 Saek la:mB2 ke:C1 pʰapDS2 tʰopDS2 taŋC1 ʔɔtDS1 ʔɯǝpDS2 dɔ:ŋA1 ma:B1 takDS1 se:B2 pʰrawA1 juəŋC1 da:tDL1 luəkDL2 ruŋA1 nɯŋC1 356 Gloss 673 extinguish, to 674 roof, to 675 cover (with cloth), to 676 deceive, to 677 buy, to 678 sell, to 679 borrow, to 680 defeat, to; be defeated, to 681 have, to 682 take, to 683 obtain, to 684 pick, to 685 steal, to 686 give, to 687 dust off, to 688 sweep, to 689 wipe, to 690 rinse, to 691 wash (clothes), to 692 wash, to 693 bathe, to PT *ɗapD *mwuŋA *hɤmB *bra:ŋA *z.ɟɯ:C *p.qa:j *ˀjɯ:mA Siamese dapDS1 muŋA2 homB1 pʰra:ŋA2 sɯ:C2 kʰa:jA1 jɯ:mA1 *be:C pʰɛ:C2 *mi:A *ʔawA *ɗajC *kepD *C̬ .lakD *haɰC *pacD *kwa:tD *ɟetD *C.lwɤ:ŋC *zakD *za:wA *ʔa:pD mi:A2 ʔawA1 da:jC1 kepDS1 lakDS2 hajC1 patDS1 kwa:tDL1 cʰetDS2 la:ŋC2 sakDS2 sa:wA2 ʔa:pDL1 Sapa dapDS1 muŋA2 pa:ŋA2 sɯC2 xa:jA1 miA2 ʔawA1 dajC1 lakDS2 haɰC1 la:ŋC2 sakDL2 ʔa:pDL1 Bao Yen dapDS1 muŋA2 homB1 pja:ŋA2 tʰɯ:C2 kʰa:jA1 pɛ:C2 ʔawA1 dajC1 -v kopDS1 lakDS2 hɤɰC1 Cao Bang dapDS1 muŋA2 homB1 rɯC2 kʰa:jA1 jɯmA1 Lungchow dapDS1 fuŋA2 humB1 pja:ŋA2 ɬɯ:C2 kʰa:jA1 jimA1 Shangsi dapDS1 homB1 soyC2 ha:jA1 jomA1 -v miA2 dajC1 -v kepDS1 lakDS2 hɤɰC1 patDS1 mi:A2 ʔawA1 dajC1 kipDS1 lakDS2 hɯ:C1 -v patDS1 moyA2 ɁawA1 ɁajC1 -i hoyC1 cetDS2 la:ŋC2 tʰakDS2 ʔa:pDL1 rakDS2 ra:wA2 Ɂa:pDL1 la:ŋC2 ɬakDS2 ɬa:wA2 ʔa:pDL1 ɬa:wA2 Yay dapDS1 foŋA2 sɯC2 ka:jA1 peB2 miA2 ʔawA1 dajC2 cipDS1 rakDS2 haɰC1 patDS1 kwa:tDL1 rɯəŋC2 θakDS2 ʔa:pDL1 Saek dapDS1 hamB1 sɯ:C2 kwa:jA1 pʰɛ:B2 mi:A2 ʔawA1 dajC1 kipDS1 lakDS2 hɤ:C1 pɛkDS1 kwa:tDL1 sɛtDS2 jakDS2 ʔa:pDL1 357 Gloss PT 694 teach, to *so:lA 695 rest, to *bakD 696 wait, to *cra:C 697 be left over, to *ʰlɯəA 698 leak, to *rwo:B 699 awaken (someone), to *plokD 700 wake up, to *k.tɯ:nB 701 grow, to; rise, to *ʰma:C 702 shrink, to *ʰrotD 703 disappear, to *ʰrwɤ:jA 704 die, to *p.ta:jA 705 kill, to *qa:C 706 expose to the sun, to *p.ra:kD 707 hang up, to *qwe:nA 708 hang down, to *ho:jC 709 smear, to *da:A 710 lay (a cloth, etc.) across, to *ba:tD 711 drive away, to *k.rapD 712 chase, to *lajB 713 be finished, to *le:wC 714 to commission *ʑaɰC Siamese sɔ:nA1 pʰakDS2 tʰa:C1 lɯəA1 ruəB2 plukDS1 -v tɯ:nB1 hotDS1 ha:jA1 ta:jA1 kʰa:C1 ta:kDL1 kʰwɛ:nA1 hɔjC1 tʰa:A2 Sapa tʰa:C1 huB2 pukDS1 tɯnB1 ma:C1 hutDS1 ta:jA1 xa:C1 taʔDL1 xwenA1 pʰa:tDL2 kʰapDS1 lajB2 lɛ:wC2 cʰajC2 sapDS1 lajB2 Bao Yen tʰɔ:nA1 tʰa:C1 lɯəA1 ruəB2 pjokDS1 tɯ:nB1 hotDS1 Cao Bang b̤ akDS2 tʰa:C1 lɯəA1 ruəB2 pjokDS2 tɯnB1 ma:C1 hotDS1 Lungchow ɬo:nA1 pakDS2 lɯ:B1 ɬu:B2 pjukDS1 ma:C1 Shangsi ɬønA1 pakDS2 luB1 pʰa:jA1 kʰa:C1 pʰja:kDL1 kʰɛ:nA1 -i hɔ:jC1 pa:tDL2 sapDS1 lɤjB2 lɛ:wC2 tʰa:jA1 kʰa:C1 tʰa:kDL1 kʰwɛnA1 hɔjC1 d̤ a:A2 b̤ a:tDL2 lɛwC2 ha:jA1 kʰa:C1 pʰja:kDL1 tʰa:jA1 ka:C1 tʰa:kDL1 ta:A2 lajB2 caɰC2 lewC1 -t Yay saC1 lɯəA1 roB2 pjokDS1 rɯnB1 maC1 rutDS1 rɯəjA1 ta:jA1 kaC1 ta:kDL1 venA1 pa:tDL2 lajB2 lewC2 θaɰC2 Saek sɔ:lA1 tʰa:C1 luəA1 rɔ:B2 plukDS1 -v rɤ:jA1 pra:jA1 ka:C1 pra:kDL1 ve:nA1 tʰa:A2 lajB2 358 Gloss 715 entrust, to I. Miscellaneous 716 1s pronoun (weak) 717 1s pronoun (strong) 718 2s pronoun (weak) 719 2s pronoun (strong) 720 3s pronoun (weak) 721 3s pronoun (strong) 722 1p pronoun 723 one 724 two 725 three 726 four 727 five 728 six 729 seven 730 eight 731 nine 732 hundred; string, to 733 single, only one PT *ʰwa:kD Siamese fa:kDL1 Sapa faʔDL1 Bao Yen Cao Bang Lungchow pʰa:kDL1 pʰa:kDL1 pʰa:kDL1 Shangsi Yay Saek *ku:A *kawA *mɯŋA *maɰA *mɯnA *manA *rawA *nɯ:ŋB *so:ŋA *sa:m *si:B *ha:C *krokD *cetD *pe:tD *kɤwC *ro:jC *ɗiəwA ku:A1 kuA1 mɯŋA2 mɯŋA2 manA2 rawA2 nɯŋB1 -t sɔ:ŋA1 sa:mA1 si:B1 ha:C1 hokDS1 cetDS1 pɛ:tDL1 ka:wC1 rɔ:jC2 diəwA1 minA2 hawA2 nɯŋB2 soŋA1 sa:mA1 siB1 ha:C1 hukDS1 citDS1 petDL1 kawC1 hojC2 kɤwA1 kɤwA1 mɤɰA2 manA2 rɤwA2 nɤŋA1 -t tʰɔ:ŋA1 tʰa:mA1 tʰi:B1 ha:C1 sokDS1 cetDS1 pɛ:tDL1 kɤwC1 rɔ:jC2 diəwA1 ɬɔŋA1 ɬa:mA1 ɬiB1 ha:C1 sokDS1 cetDS1 pɛtDL1 kawC1 dɛwA1 kawA1 maɰA2 mɤnA2 lawA2 nɤŋA1/B2 ɬo:ŋA1 ɬa:mA1 ɬi:B1 ha:C1 hukDS1 citDS1 pe:tDL1 kawC1 ɬo:jC2 de:wB1 kawA1 maŋA1 -t manA1 -t ɬø/ŋA1 ɬa:mA1 ɬoyB1 ha:C1 lokDS2 citDS1 petDL1 kawC1 kuA1 kuA1 mɯŋA2 manA2 rawA2 θoŋA1 θa:mA1 θiB1 haC1 rokDS1 satDS1 petDL1 kuC1 rojC2 dewA1 ku:A1 ku:A1 mɯŋA2 manA2 ro:A2 -v nɯŋB2 sɔ:ŋA1 sa:mA1 si:B1 ha:C1 rɔkDS1 cɛtDS1 pe:tDL1 ku:C1 rɔ:jC2 diəwA1 359 Gloss 734 pair 735 how many, several 736 all 737 many, much 738 little, few 739 one or so, any 740 half 741 each other, together 742 other 743 measure from thumb to fingertip 744 bite (n.), speech 745 classifier for things 746 cord, string 747 classifier for long, thin, object 748 classifier for tools 749 litter (of young) 750 time (classifier) 751 name of first month 752 last (year) 753 daytime PT *gu:B *ki:C *daŋA *ʰla:jA *ʰno:jB *sakD *grɤŋB *kanA *ʔɯ:nB *ɣɯ:pD *gamA *ʔalA *sa:jA *selC *mwa:kD *gro:kD *bajA *ciǝŋA *kla:jA *ŋwanA Siamese kʰu:B2 ki:B1 -t tʰaŋ4 la:jA1 nɔ:jB1 sakDS1 kʰrɯŋB2 kanA1 ʔɯ:nB1 kʰɯ:pDL2 khamA2 ʔanA1 sa:jA1 senC1 kʰrɔ:kDL2 kla:jA1 wanA2 Sapa kiC1 tiŋA2 -v la:jA1 kanA1 xɯpDL2 kamA2 sa:jA1 ciəŋA1 ka:jA1 vinA2 -v Bao Yen ki:B2 -t taŋA2 la:jA1 Cao Bang g̈ uB1 kiC1 d̤ aŋA2 la:jA1 Lungchow ku:B2 ki:C1 taŋA2 la:jA1 Shangsi koyC1 la:jA1 tʰakDS1 ɬakDS1 ɬakDS1 Yay kuB2 ciC1 la:jA1 nojB2 -t kanA1 ʔɯ:nB1 kanA1 kanA1 ʔɤnB1 kamA2 ʔanA1 tʰa:jA1 g̈ amA2 ɁanA1 ɬa:jA1 ʔanA1 ɬa:jA1 ɁanA1 ɬa:jA1 hupDS2 θa:jA1 ma:kDL2 pajA2 ciəŋA1 ca:jA1 wanA2 ma:kDL2 ciǝŋA1 ca:jA1 wanA2 ma:kDL2 kjo:kDL2 pajA2 ci:ŋA1 kja:jA1 vanA2 pajA2 ciŋA1 ka:jA1 ɲɤnA2 -v fa:kDL2 cokDL2 vanA2 Saek ki:C1 tʰɯŋA2 -v la:jA1 nɔ:jB2 -t sakDS1 tʰrɯŋB2 kinA1 -v ʔɯ:nB1 (kʰɯ:pDL2 -i -v) ʔalA1 sa:jA1 sɛlC1 ma:kDL2 rɔ:kDL2 ciəŋA1 tla:jA1 ɲɛnA2 360 Gloss 754 night 755 day 756 early morning 757 night 758 late at night 759 early meal 760 late meal 761 evening meal 762 day before afterday 763 yesterday 764 tomorrow 765 day after tomorrow 766 year 767 upstream, above 768 downsteam, below 769 below 770 inside 771 outside 772 middle 773 side 774 this PT *ɣɯ:nA *mwɯ:C *ɟawC *ɣamB *ɗɯkD *ŋa:jA *C̬ .lwi:ŋA *ɟm̩ .rawA *zɯ:nA *ŋwa:A *ɟm̩ .rukD *C̬ .rɯ:A *pi:A *ʰnɯəA *taɰC *lwɤ:ŋB *C̥ .daɰA *l̩ .no:kD *kla:ŋA *C̥ .bɯǝŋC *najC Siamese kʰɯ:nA2 mɯ:C2 cʰa:wC2 kʰamB2 dɯkDS1 ŋa:jA2 lɛ:ŋA2 pʰrawA2 sɯ:nA2 wa:nA2 pʰruŋB2 -f rɯ:nA2 -f pi:A1 nɯəA1 ta:jC1 la:ŋB2 najA2 -i nɔ:kDL2 kla:ŋA1 bɯǝŋC1 ni:C2 -v Sapa xɯnA2 mɯC2 cawC2 xamB2 dɯkDS1 ŋa:jA2 leŋA2 sɯnA2 ŋwa:A2 puʔDL1 hɯA2 piA1 taɰC1 la:ŋB2 noʔDL2 ka:ŋA1 niC2 Bao Yen kʰɯnA2 mɯ:C2 cawC2 kʰamB2 ŋa:jA2 lɛ:ŋA2 lɛ:ŋA2 tʰɯ:nA2 ŋwa:A2 pjukDL2 rɯ:A2 pi:A1 nɯəA1 taɰC1 la:ŋB2 nɔ:kDL2 ca:ŋA1 bɯəŋC1 najC2 Cao Bang g̈ ɯnA2 mɯA2 g̈ amB2 lɤkDS1 -i ŋa:jA2 lɛŋA2 b̤ jɤwA2 rɯnA2 ŋwa:A2 lɯA2 nɯəA1 tɤɰC1 dɤɰA1 nɔkDL2 bɯəŋC1 nɤjC2 Lungchow kɤnA2 cawC2 kamB2 ŋa:jA2 le:ŋA2 pjawC2 ɬinA2 va:A2 pjukDS2 lɯ:A2 pi:A1 nɯ:A1 taɰC1 la:ŋB2 daɰA1 no:kDL2 kja:ŋA1 bɤ:ŋC1 najB2 Shangsi hanA2 sawC2 hamB2 ŋa:jA2 lɤŋA2 ɬonA1 tokADL2 loyA2 poyA1 tajC1 -v doyA1 nokADL2 ka:ŋA1 noyC2 Yay hɯnA2 fɯC2 sawC2 hamB2 dakDS1 ŋa:jA2 riŋA2 sawA2 sokDL2 rɯA2 piA1 daɰA1 rokDL2 ca:ŋA1 niC2 -v Saek ɣɯnA2 mɯ:C2 ɣamB2 dɯkDS1 ŋa:jA2 lɛ:ŋA2 tʰɔ:k5 rɯ:A1 pi:A1 tɤ:C1 luəŋB2 rɤ:A1 lɔ:kDL2 tla:ŋA1 viǝŋC1 ni:C2 -v 361 Gloss 775 before 776 tip, end 777 tip, highest point 778 edge 779 gradually, slowly 780 not (strong 1) 781 not (strong 2) 782 not (weak) 783 not yet 784 do not 785 which 786 also 787 with, and 788 matter PT *ko:nB *pla:jA *ɲo:tD *ri:mA *gɔ:jB *ɓawB *bo:B *mi:A *pajB *ʰɲa:B *ɗaɰA *ko:C *kapD *ɣwa:mA Siamese kɔ:nB1 pla:jA1 jɔ:tDL2 rimA2 kʰɔjB2 bo:B1 miA2 pajB1 ja:B1 dajA1 kɔ:C1 kapDS1 kʰwa:mA2 Sapa kɤnB1 ɲotDL2 kojB2 boB1 daɰA1 koC1 Bao Yen kɔ:nB1 pja:jA1 ɲɔ:tDL2 kɔ:jB2 bɤwB1 mi:A2 pajBA1 dɤɰA1 Cao Bang kɔnB1 pja:jA1 ɲɔtDL2 rimA2 g̈ ɔjB2 bɤwB1 mi:C1 -t pajB1 dɤɰA1 Lungchow ko:nB1 pja:jA1 jo:t6 ko:jB2 bawB1 bo:B1 bo:B1 miB1 -t ja:B1 naɰA2 -i Shangsi kønB1 miC1 -t kapDS1 kapDS1 wa:mA2 va:mA2 Yay konB1 pja:jA1 Saek kɔ:nB1 pla:jA1 ɲu:tDL2 -v rimA2 bawB1 boB1 boB1 miA2 bo:B1 bo:B1 jiəB1, ja:B1 koC2 -t ja:B1 dɤ:A1 kapDS1 Notes to glossary 1. The reflexes of this etymon in NT dialects point to PT *-r. It is generally thought to be the same etymon as 'hair knot' found only in CT and SWT with *-l-. The Siamese form generally refers to 'hair knot' but means 'head' in the expression /puətDL1 siənA1 wiənA2 kla:wC1/ 'to have a headache, to be confused'. See 'hair knot'. 2. cf. 頭 tóu ‘head’ (Sagart 1999) from MC dəu, LH do, and OC *dô. This etymon is not found in NT, and might in fact be a post-PT loanword. 5. Reflexes in some dialects, e.g. Ningming /honA1/, point to an immediate stage *χ-. See 'leg'. 6. The Siamese form means 'nose brigde'. 11. cf. PAN *matá ‘eye’. 14. The Cao Bang form is found only in compounds, e.g. /kʰi:C1 vanA2/ 'food stuck between teeth'. 15. cf. 顎 è ‘palate’ (Schuessler 2007: 224) from MC ŋâk, LH ŋɑk, and OC *ŋâk. 16. NT varieties point to earlier *ml-. The cluster is probably due to influence of the final *-m in PT *C̬ .namC 'water', with which the etymon 'saliva' usually occurs, cf. Saek /namC2 mla:jA2/ 'saliva'. 18. The initial *k.- may have been a prefix that was lost early on in dialects that have reflexes with tone A2. 19. This etymon independently came to be the general word for 'tooth' in most dialects, e.g. Lao, White Tai, Lungchow, Cao Bang, etc. Siamese still preserves the original meaning. See 'tooth'. 20. Shangsi shows /kʰ-/ instead of the expected /h-/. 22. cf. 喉 hóu ‘throat’ (MC ɣəu; LH go; OC *gô). 24. cf. PAN *qabáɣa ‘shoulder’. 25. cf. 肩 jiān ‘shoulder’ (MC kien; LH ken; OC *kên). 27. The /-ŋ/ in Yay and other NT languages resulted from a sporadic nasalization, i.e. *mwɯ:A >*mwɯ̃ :A > *mwɯ:ŋA. 28. Not found in NT. 29. The long vowel in Saek seems irregular but the Lakkja cognate /pli:pDL1/ also has a long vowel. 362 30. The Yay form means 'a piece or section of a tree cut' and might not be related. 31. Some dialects including Wuming, Yongbei, Shangsi, and Ningming point to an immediate stage *χ-. 32. Saek /k-/ is aberrant. It might be a result of contamination from 'leg'. 33. cf. 脛 jìng ‘shin’ (MC ɣieŋC; LH geŋC; OC *gêŋh). Lungming unexpectedly has /k-/. 38. cf. 腰 yāo ‘waist’ (MC ʔijäu, LH ʔiɑu, OC ʔiaw). 39. PT *C.wɯǝtD > *hwɯǝtD. 41. Sapa /dɯA1/ is found in /dɯA1 majC2/ 'wood knuckle' only. See 'gall bladder, bile'. 49. cf. 尿 niào ‘urine’ (Schuessler 2007: 401)from MC nieuC, LH neuC, and OC *niâu(k)h. Independently in SWT and NT dialects, *n- became palatalized in front of *-i-, yeilding *ɲiəw. NT further simplify the intermediate form to due *ɲɤw, i.e. *niəw > *ɲiəw > *ɲɤw. CT /new/, /ne:w/ and /nɛ:w/ most likely resulted from simplification of *iəw to *e:w, cf. 'single, one', but it is also possible that they are late Chinese loans, cf. Mandarin niao. 53. In many dialects including Shan, Yay, Saek, etc., this etymon means 'brain'. In other dialects, it also refers to 'bone marrow', cf. Bao Yen /ʔɔ:kDL1 du:kDL1/ 'brain'. See 'brain'. 54. The onset reflexes in Yay and other NT dialects show sporadic nasalization under the influence of *-n. 57. Contrast PT *n.m- > *mn- > *ml- in NT with PT *n.m- > *nɯ- in SWT and CT. 59. Cao Bang shows PT *r.t- > *tr- > /th-B1/ but Shangsi indicates *r.t- > *d- > /t -B2/. Standard Siamese now shows an irregular /ŋ-/, possibly due to spontaneous nasalization. 60. cf. 膿 nóng ‘pus’ (MC nuoŋ; LH nouŋ; OC *nǔŋ). 63. The /-l-/ in Siamese /khlajA2/ is unexpected. CT forms /lajA2/ or /jajA2/ are not related to this etymon. 64. In many dialects including White Tai, Black Tai, Lue, Leiping, etc., this etymon means 'brain'. In other dialects, it occurs in a compound with *ʔwu:kD 'bone marrow', cf. Bao Yen /ʔɔ:kDL1 ʔɛ:kDL1/ 'brain'. See 'bone marrow'. 66. cf. 肺 fèi ‘lung’ (MC pʰjwɐiC, LH pʰuɑs, OC *phats). 67. Lakkja /kja:jC1/ 'intestine' suggests PT *k.sajC, but Tai-internal evidence is lacking. 363 70. Cao Bang /d-/ might have been due to contamination from 'navel (2)'. The confusion between 'navel' and 'gall bladder, bile' is common among Tai dialects, cf. Sapa /sajC1 biA1/ 'navel, umblilical cord' from PT *sajC ‘intestine’ + *ɓli:A ‘gall bladder, bile’. 73. The irregular tonal reflexes might be related to the taboo nature of this etymon. 75. It is conceivable that this etymon was in fact *q.ma:A, cf. Lakkja /kʰwõA1/, but Tai evidence is lacking, 76. It is possible to reconstruct *q.mu:A, for this etymon, cf. Lakkja /kʰũA1/, but Tai-internal evidence is lacking. 79. cf. 馬 mǎ ‘horse’ (MC maB; LH maB; OC *mrâʔ). 80. cf. PAN *tumáy ‘bear’. 82. This etymon is reconstructible at the PT, though ultimately from 象 xiàng ‘elephant’ (MC zjaŋB; LH ziɑŋB; OC *s-jaŋʔ). Wuming /ɕiəŋC2/, Yongbei /tsɯ:ŋC2/, Lianshan /θe:ŋC2/ among others are later Mandarin loan from the same Chinese etymon. 83. PT *kwɯəŋA > *wɯəŋA > /vɯəŋA2/ in Yay and Saek. 87. The Saek form /pa:ŋB1/ is unrelated. 90. PT *ʰɲ- became *h- in front of front vowels in all SWT and some CT dialects. 92. Thai form /nimB2/ is not related 93. See 'tree, wood'. 95. The vowel irregular vowel reflexes in SWT and many CT dialects suggest that this etymon might in fact be a post-PT loan. The unaspirated /p-/ indicates an early date of borrowing (see §3.5) 96. cf. 雁 yàn ‘wild goose’ (MC ŋanC; LH ŋanC; OC *ŋrans). 97. cf. PAN *manúk ‘bird’. 99. cf. 燕 yàn ‘swallow’ (MC ʔienC; LH ʔenC; OC *ʔêns). 100. See 'tree, wood'. 104. The Yay form /θajA1/ is not related and might have come from Chinese 螄 sī ‘snail’. Chongzuo has both /ɬai1/ and /ho:i1/. 106. NT dialects point to *B. It is hard to determine what the original tone was. 364 107. PT *ŋiəwA > *ɲiəwA > Lungchow /ji:wA2/. 109. Wuming has /klweC1/. For CT, see Jingxi and Debao /kweC1/. 111. Saek /rɔ:A2/ is not related. It is a borrowing from Vietnamse rùa 'turtle'. 113. Lingyue /ŋwaA2/ attests PT*-w-. 114. cf. 鱷 è ‘crocodile’ (Schuessler 2007: 223) from from MC ŋâk, LH ŋɑk, and OC *ŋâk. This etymon came to mean 'mythical water creature' in SWT and many CT dialects. 115. This word might have in in fact been *l.ta:kD in PT, as suggested by Lakkja /la:kDL2/ 'land leech', but Tai-internal evidence is lacking. This etymon is likely to be related to PAN *qaɬimátek and *qaɬimatéq, both meaning 'leech'. 122. cf. 螟蛉 míng-líng ‘insect’. See §4.4.5 and Pittayaporn (to appear-b) for discussion of the Siamese reflex. 123. cf. Proto-Semai *smɤ:c. See discussion in §5.5. 127. Saek /lɯəp6/ is loan from Siamese or Lao. 130. This etymon might have in fact been *q.matD in PT, *q- being an animal suffix, cf. Lakkja /kʰwõtDS1/ 'flea'. Tai evidence is lacking. See 'dog' and 'pig'. 132. The medial -w- is preserved in some NT dialects including Huanjiang /rwiA2/ and Nandan /lwejA2/. 133. Wuming /klwa:wA1/ and Liujiang /kjwa:wA1/, among others, preserve the medial *-w-. 141. cf. 角 jiǎo ‘horn’ (MC kåk; LH kɔk; OC *krôk). Bao Yen /s-/ suugests PT *qr- but other dialects all point to *q-. 142. cf. 牙 yá ‘tusk’ (MC ŋa; LH ŋa; OC *ŋrâ). 146. Aberrant final /-p/ in some CT and NT dialects, e.g. Yay /cip3/, Leiping /kipDS1/, and Ningming /kipDS1/. Also see 'mushroom', 'tick', and 'hail'. 148. See 'tree, wood'. 365 149. Bao Yen, Cao Bang and, and a few others show reflex of *p.t-, but Leiping, Lungming, and Saek, among others, show reflex of PT *t-. In this latter set of dialects, the onset might be due to contamination from *tapD 'liver', cf. Siamese expression /tapDS1 tajA1/ 'internal organs'. 150. Sporadic nasalization in many varieties including Siamese. 151. PT *ɯəj simplified to /ɯə#/ in many dialects including Siamese. 153. The Siamese form /najA2/ is aberrant, possibly due to dissimilation from /lekDS2/ 'iron, metal', i.e. *lekDS2 lajA2 > /lekDS najA2/. 154. Several high-frequency words in Bao Yen and Cao Bang have /aj/ and /aw/ instead of the expected /ɤj/ and /ɤw/. 162. Not found in NT 164. The *q.-, is tentatively reconstructed to account for the /h-/ in Yay. 170. Not found in SWT. For CT, Debao, Jingxi, Yanshan Nung, and Guangnan Nung have /ra:ŋA2/. 173. cf. 稼 jià ‘sow grain’ (MC kaC; LH kaC; OC *krâh). 183. I tentative reconstruct *ɢw- to distinguish it from *w-, which would give /fa:ŋA2 / in Siamese and Lungchow. 187. Siamese and Lao /sajA2/ is a Mon-Khmer borrow, cf. Nyakur /chrɛ̤ y/ (Shorto 2006). 190. cf. 芥 jiè ‘mustard plant’ (MC kăiC; LH kɛs; OC *krêts) 191. Proto-Kra has *məlA (Ostapirat 2000) suggesting PT *malA but Saek has /-n/, possibly an influence from Lao or Siamese. 195. cf. PAN *leŋa ‘sesame’. 196. Possibly a late borrowing from Chinese 茶 ‘tea’ (MC ḍa; LH ḍai; OC *dra < *r-lâ). The form /cɛB2/ in Black Tai, and White Tai, as well as /cɛ:A2/ in Bao Yen is from Vietnamese trè 'tea'. 198. cf. PAN *quwai ‘rattan’. 202. The Saek form /hetDS1/ is a Siamese/Lao loan because both the onset and the vowels are unexpected. Ningming has /ɬapDS1/ with an aberrant /-p/. 203. cf. 蕨 jué ‘fern, bracken’ (MC kjwɐt; LH kyɑt; OC *kot). 206. cf. PAN *daɬúm 'water' (Blust's *daNum). 366 207. cf. Proto-Malayo-Polynesian *apuy < PAN *sehapúy ‘fire’. 208. cf. 熛 biāo ‘leaping flame’ (Schuessler 2009) from MC pjiäu , LH piau, and OC *piau. PT *-e:w became PSWT *-ɛ:w, which is now reflected as /-e:w/. See 'liquid'. 210. cf. 煤 méi ‘soot, coal’ (author’s etymology) from MC mwậi, LH mə, and OC *mə̂ . 212. Saek /luəA1/ has an irregular rime, and is likely a borrowing from Lao. 214. Li (1977: 95) cites Lungchow /pjawB2/ as reflex of this etyma but this form in fact came from a different etymon— *brawB ‘leaf ashes’. Debao have both /tawB2/ from *dawB and /pjawB2/ from *brawB. 216. Not found in SWT and CT. 217. Rong’an has /kwɤA1/ attesting the medial *-w-. 219. Proto-Kra has *(m-)ɳəlA ‘fat’(Ostapirat 2000) suggesting PT *malA but Saek has /-n/, possibly an influence from Lao or Siamese. 220. The Siamese. Cao Bang, and Lungchow forms mean 'mud'. 221. Not found in CT. 222. This is most likely a relative early Chinese loan from before the Later Han period, cf. 沙 shā ‘sand’ (MC ṣa; LH ṣa < ṣai; OC *srâi). 225. Some NT and CT dialects have /uə/. 228. Comparing the vowel with Chinese forms, this etymon may have been borrowed from Chinese relatively recently, cf. 地 dì ‘earth, ground’(MC diC; LH diC; OC *r-laih) 231. The reconstruction of the onset is tentative because crucial evidence from NT is lacking. 234. Bao Yen /s-/ suugests PT *-r- but all other dialects indicates that PT did not have a liquid in this etymon. 236. cf. Vietnamese suối 'mountain stream'. Longsheng has /jwi/ from earlier *rwiC. 243. This etymon was the general PT word for 'star'. 244. This etymon is normally used together with *t.na:wA 'star' to refer to stars in the sky, in contrast with falling stars, cf. Bao Yen /da:wA1 di:B1/ 'star in the sky' vs. /da:wA1 ha:ŋA1/ 'falling star', /ha:ŋA1/ being Bao Yen's reflex for 'tail'. 248. cf. 霧 wù ‘fog’ (MC mjuC; LH muoC; OC *moh). 367 250. Cao Bang has /bɤnA1/ for 'heaven', and /va:C2/ for 'sky' and 'weather'. The Siamese form means 'on, above' 251. See 'sky, heaven'. 254. Tianlin has /mja:pDL2/. 257. Among NT dialects, Longsheng and Wuming have /rwa:jA2/ and /ra:iA2/ respectively. 259. The short vowel in Siamese is unexpected. The NT and CT/SWT may in fact be unrelated. 260. cf. 陰 yīn ‘overcast, shade’ (author's etymology) from MC ʔjəm, LH ʔɨm, and OC *ʔəm (Baxter’s *ʔ(r)jum). Saek has /jumA1/ but this is probably not related. 262. Saek /lɛ:ŋC2/ is probably a borrowing from Siamese or Lao. The expected vowel reflex is /e:/. 265. NT dialects have *C due to contamination from *nwo:ŋC 'younger sibling'. 267. For NT, Tianlin, Lingyue, Pingguo, Hechi, etc., all have /pawB1/. 269. The SWT form is contaminated by the post-PT etymon *ja:B 'maternal grandmother'. See Li (1971) and Matisoff (1992) for discussion). 273. SWT languages have *A probably due to contamination from *hla:nA 'grandchild'. Saek /le:nA1/ is a Siamese borrowing. 277. Lungchow and Shangsi show reflexes of *B2, perhaps due to influence of *me:B or *bo:B. 278. Siamese has an archaic form /ʔa:wA1/. 282. Kosaka (2007) claims that the irregular Saek onset reflex results from contact with *lɯ:kD 'child', i.e. *lɯ:kD baɰC > (*lɯ:kD gbwaɰC) > *lɯ:kD gwaɰC > Saek /lɯkDS2 khwɤ:C2/. 284. The final /-ŋ/ in Sapa and Bao Yen results from an assimilation to the onset of *no:jC 'small', cf. Sapa /diŋA1 nojC2/ 'child'. The Saek form is /dɛkDS1/ is loan from Siamese or Lao. 286. The Siamese form now means 'servant'. 289. The simple onset suggests that this etymon might in fact be a relatively recent loan, cf. 娘 niáng ‘young woman, mother’ from MC nrɨaŋ (Pulleyblank 1991) 291. NT points to *B, probably due to contamination from PT *me:B 'mother' or post-PT *ja:B 'lady', cf. Siamese /mɛ:B2 ma:jC1/ 'widow' and Yay /ja:B2 ma:jB1/ 'widow'. 368 293. cf. 字 zì ‘polite name, written character’ (MC dzɨC; LH dziəC; OC *dzəh). 294. cf. 主 zhǔ ‘host, master, owner’ (author’s etymology) from tśjuB, LH tśjoB, OC toʔ. Saek also has /ca:wC1/, a borrowing from Siamese or Lao. 295. SWT and CT dialects point to *C, probably due to contamination from /kra:C/ 'slave'. 298. cf. 巫 wū‘shaman’ (MC mju, LH muɑ, OC *ma). 299. The Saek form /pʰi:A1/ is a borrowing from Thai or Lao. 300. Cao Bang /ma:ŋA2/ means 'coffin for dead body'. 301. cf. 魂 hún ‘spiritual soul’ (MC ɣwən, LH ɣuən, OC *wə̂ n). 306. Saek has /ja:A1/ but this is most likely a Simese/Lao loan. 312. The fricative *ʑ is tentative. 315. Wuming has two forms for ‘pillar’: native Tai /θaw A1/, and a recent loan /ɕaɰA2/ from Chinese 柱 zhù ‘post, pillar’. 319. cf. 凳 dèng ‘stool, bench’ from MC təŋh. This etymon might in fact be a relatively late borrowing. 327. This etymon is related to PAN *panaq ‘crossbow’, Chinese 弩 nǔ ‘crossbow (MC nuoB, LH nɑB, OC *nâɁ), as well as Mon-Khmer *snaʔ (Shorto 2006). The medial *-w- perhaps suggests earlier *p.na:C. 329. PT *i:w became /iəw/ in Yay. See 'to carry (a bag)'. 330. For NT, Lingyue /hɔA1/. 331. cf. 鉗 qián ‘pincers, thongs’ (MC gjäm; LH gɨam; OC *gam). This etymon might in fact be a post-PT borrowing. 334. This etyma may have functioned both as a noun 'carrying pole' and a verb 'to carry with a carrying pole', cf. Siamese /ha:pDL1/. 337. Wuming has /ra:ŋB1/ instead of the expected /kla:ŋB/. 338. cf. 盎 àng ‘bowl’ (MC ɁâŋC; LH ɁɑŋC; OC *Ɂâŋh). 342. The reconstruction of the onset is tentative because crucial evidence from NT is lacking. 344. The medial *-w- is attested as /-u-/ in Lingyue /luəjA1/ and Nandan /ruəjA1/. 369 345. cf. 書 shū ‘written document’ (MC śjwo; LH śɑ; OC *lha). 348. cf. 徽 huī ‘rope’ (Schuessler 2007: 286) from MC xjwei, LH hui, OC *hməi. Saek /mi:A1/ is not related. 353. cf. 藍 'indigo, blue'. This etymon refers to Indigofera tinctoria .Saek form /kha:mA2/ is a loan from Lao or Thai. NT points to *ɟa:mA. 354. Baphicacanthus cusia (Ness) Bremek 356. cf. 鍼 zhēn ‘needle’ (Schuessler 2007: 610) from MC tśjəm, LH tśim < kim, OC *kim. 359. cf. 笠 lì ‘bamboo hat’ (Schuessler 2007: 351) from MC ljəp, LH lip, and OC *rəp (Baxter’s *g-rjəp). 365. Possibly related to 園 yuán ‘garden’ (author's etymology) from MC jwän, LH wan, and OC *wen. 370. The *-w- is preserved in Tianlin Rong'an /ɕwajA1/ and Rong'an /kjwajA1/, among others. 371. cf. 軛 è ‘yoke’ (MC Ɂɛk; LH Ɂɛk; OC Ɂrêk). 378. Sapa, Phu Thai, and many SWT dialects have /ɯəŋ/ instead of the expected /a:ŋ/. 379. Not found in NT 380. Shan dialects have /kʰ-/ instead of the expected /h-/. 383. Saek /ha:ŋA2/ is a borrowing from Lao. 386. cf. 簰 ‘raft’(MC baɨ, LH, bɛ, OC *brê). 387. Liujiang /hjwanA1/ and Huanjiang /wanA1/ attests the medial *-w-. 388. The medial *-w- is attested in Tianlin and Donglan /lwiA2/. 389. cf. 鞍 ān ‘saddle’ (MC Ɂân; LH Ɂɑn; OC *Ɂân). 392. cf. PAN *tidem 'dark, black' 397. The Siamese form means 'not bright, dusty'. 398. Black Tai and Lue show reflexes of *ɗ-. 403. cf. 肥 féi ‘fat’ (MC bjwei; LH bui; OC *bəi). PT *-w- is preserved in Huangjiang and Rong'an /pwiA2/, among others. 370 407. Not found in CT and SWT 408. The tonal irregularity in certain dialects due to confusion with *ʰno:jB 'few, little'. Siamese and Lungming still maintain the distinction. 412. This form might in fact be reconstructed as *c.na:A, as suggested by Lakkja /cã:A1/, but Tai-internal evidence is not available. 417. The reconstruction of the onset is tentative because crucial evidence from NT is lacking. 418. Guigang /ʔwunB1/ attests the medial *-w419. cf. PAN *qudip ‘alive’. 422. Some CT languages, e.g. Lungchow, Leiping, and Pingsiang, point to PT *ɗ-. These foems might not be related. 423. SWT varieties and Sapa abberrantly show reflexes of PT *uəm. 427. Qinzhou has /ɬa:C2/. 436. cf. PAN *leqacém ‘sour’. 442. Yongnan has /he:wA1/. The vowel in Siamese is irregular probably due to false etymological connection to 'green'. 447. This etymon shows a sporadic change from *-ɤj > PSWT *-i:. 452. See Gedney (1997) for discussion. 453. It is not clear why many dialects show /-w-/. See Gedney (1997) for discussion. Perhaps, this etymon goes back to PT *kl̩ .wajA. 454. Saek form is irregular. This might be a case of spontaneous nasalization, or contamination by *ʰŋa:jA 455. This might not be reconstructible (see §3.6.7.4) 456. The expected vowel reflex in Saek is /e:/. The forms /lɛmA1/ and /lɛ:mA1/ are most likely loans from Siamese or Lao. 460. Although *e: usually gives /ɛ:/ in Siamese, but *e:w gives /e:w/. See 'flame' and 'waist'. 462. The Saek form is from a case of dialect mixing with Siamese or Lao. 463. The Saek form is a Siamese/Lao borrowing. 465. Wuming has /tumA2/. 470. It is possible to reconstruct *r.mi:A for this etymon, cf. Proto-Kra (Ostapirat 2000) *r-meA, but Tai-internal evidence is lacking. 371 472. It is not clear whether the original PT tone was *B or *C. 474. The Saek form /phɛ:ŋA2/ is a borrowing from Siamese or Lao as suggested by the vowe /ɛ:/. 486. Not found in NT. 487. Not found in NT. 488. Not found in CT and SWT. 490. This etymon is found in Siamese as petrified part of the noun /sa'ʔɯkDS1/ 'hiccups'. See Kullavanijaya (1992). 494. In Siamese, this etymon has become grammaticalized to an adverb 'not yet'. 497. cf. Chinese 染 rǎn ‘to dye’ (Schuessler 2007: 239) from MC ńźjämB/C, LH ńamB/C, and OC *namʔ/ *nams, and Vietnamese nhuộm 'to dye'. 499. cf. 繡 ‘to embroider’ (MC sjəuC; LH siuC; OC *siu(k)h. For NT, see /ɬe:wB1/ in Tiandong, and /θe:wB/ in Tianlin, Du'an, etc. 501. cf. Chinese 播 bò ‘to sow, broadcast’ (Maspéro 1912) from MC pwâ C, LH pɑi C, and OC *pâih. Vietnamese vãi 'to broadcast (rice)'. 502. cf. PAN taɬem (Blust's *taNem) 510. cf. 庠 xiáng ‘to provide for’ (Schuessler 2007: 532) from MC zjaŋ, LH ziɑŋ, and OC *s-jaŋ or *s-laŋʔ. 511. cf. 養 yǎng ‘to nourish’ (Schuessler 2007: 558-559) from MC jiaŋB, LH jɑŋB, and OC *jaŋʔ or *laŋʔ. 512. This etymon means 'lightning' in CT dialects. 516. cf. 稱 chēng ‘to weigh’ (MC tśʰjəŋ; LH tśʰɨŋ, OC *k-hləŋ), or chèng ‘equal to’ (MC tśʰjəŋC; LH tśʰɨŋC, OC *k-hləŋh). 517. cf. Chinese 報 bào ‘to announce’ (MC pâuC; LH pouC; OC *pûh), and Vietnamese bảo 'to announce, to inform'. 518. Saek /ŋ-/ is perhaps due to spontaneous nasalization. 520. Leiping /ho:A1/ and Lungming /ho:A1/ are irregular. 524. cf. PAN *táŋic 'to weep'. 525. For NT, Laibin /kɣa:ŋA2/, Liujiang /kja:ŋA2/, etc. 526. The Saek form /hawB1/ is probably a loan from Lao or Thai that replaced the expected *rawB1. 372 531. The reconstruction of *-ɲ is tentative. See §5.5. 533. The medial *-w- is attested in many NT dialects, i.e. Liujiang /hjweA2/, Hechi /rwiA2/ etc. 541. Western Nung has /ŋa:wA1/ in /ŋa:wA1 ŋa:pDL2/ 'to yawn'. The nasal is probably due to contamination from /ŋa:pDL2/. The dialect also has /cʰa:wA1/. 542. The Saek reflex points to PT *b- rather than *p-. 543. The Siamse aberrantly have / -r-/ instead of the expected /-l-/. 545. The Saek form /kanC1/ is a Lao loan. 549. NT dialects including Wuming and Po-ai have /nɯnA2/ point to PT *ɯ. 553. Most NT dialects have this etymon, even though Yay and Saek do not, e.g. Wuming, Lingyue, Laibin /ko:pD/. 561. PT *i:w became /iəw/ in Yay. See 'chisel'. 562. The *-w- is preserved in Tianlin Rong'an /lwa:mA1/ and /jwa:mA1/ among others. 563. Lakkja has /kja:kD2/, possibly suggesting *g.-. 567. Saek has /ɲi:C2/, and Yay has /jiC2/ and /viC2/, pointing to *ɲwi:C. 574. Leiping and Lungming have /ji:pDL1/ and /ʔi:pDL1/ respectively. 582. SWT and CT dialects point to earlier voiceless onset probably due to contamination from *mɯə 'to return*. Kam-Sui languages point to voiceless initial *ʰm- (Thurgood 1988). 583. Wuming and Shangsi unexpectedly show reflexes of *i:. 584. cf. PAN *dateŋ ‘to arrive’. 588. See 'tree, wood'. 589. Etymologically related to *r.no:kD ‘outside’. 591. cf. 過 guò ‘to transgress’ (MC kwâB; LH kuɑiC; OC *kôih). 593. Saek /lɔnB1/ is most likely a borrowing from Thai/Lao 373 595. The Saek form /tʰlɤ:tDL2/ is not related. 598. Saek show /ɯ/ instead of /i/ perhaps due to the influence of both /b-/ and /-l/. 599. Wuming has both /wuənA2/ attesting the *-w-. 601. Wuming /plenA1/ is not related. 605. Phu Thai has /hwa:tDL2/. 607. This etymon might in fact be a late Chinese borrowing, cf. 浮 fú ‘to float’(author’s etymology) from MC bjəu, LH bu, and OC *bu. 610. cf. 闔 hé ‘to shut’ (author's etymology) from MC ɣâp, LH gɑp, and OC *gâp. 615. Some SWT and CT including Siamese, Black Tai, White Tai, Sapa, and Shansi, show reflexes of *-e-. 624. SWT and some CT varieties point to *p- but NT and some other CT point to *b-. 626. cf. PAN *talaw ‘to fear’. 643. cf. 剝 bō ‘to peel’ (MC påk; LH pɔk; OC *prôk). The correspondence is regular but if this is indeed a Chinese loan, the absence of *- l- suggests that it is a relatively recent loan. 653. The lack of *-r- suggests that it may in fact be a post-PT loan from Chinese, cf. 解 jiě to untie, undo’ (MC kaɨB; LH kɛB; OC *krêɁ). 658. Shanglin and Guangnan /ʔwatDS1/ attests the PT *-w-. 660. Some SWT languages, e.g. Lao, Phu Thai, etc., have forms that would point to earlier *ɓo:ŋA, which is probably not related to this etymon, cf. Kapong /dɔ:ŋA1/ 'to pickle' vs. /bɔ:ŋA1/ 'a kind of dipping sauce (not involving picking)'. 668. Shangsi /dutDL1/ means 'fever'. 669. cf. Vietnamese luộc 'to boil (tr.)'. 671. cf. 蒸 zhēŋ ‘to steam’ (author's etymology) from MC tśjəŋ, LH tśɨŋ, and OC *təŋ. 678. Some dialects including Wuming, Yongbei, Shangsi, and Ningming point to an immediate stage *χ-. See 'leg'. 679. For NT, see Qiubei /dʑəmA1/. 680. cf. 敗 bài ‘to be defeated’ (MC baiC; LH bas; OC *brats). 374 683. See 'tree, wood'. 685. Lakkja has /lakDS2/. Some NT dialects have reflexes that would point to *ɟakD, e.g. Wuming /ɕakDS2/, but this etymon is a post-PT loan, and is related to Siamese /sɯkDS1/, cf. 賊 zéi ‘thief’ (author’s etymology) from MC dzək, LH dzək, and OC *dzə̂ k. 686. Many dialects including Lungchow, Yuan etc. points to *ɯ:. This monophthongal reflex may have to do with the fact that the verb 'to give' occurs frequently and is extremely susceptible to phonological reduction and grammaticalization. 688. Bao Yen has /kwɛ:tDL1/ from Vietnamese quét 'to sweep'. 695. In many NT and CT languages including Debao, Lungchow, Wuming, Liujiang etc., this etymon has come to mean 'tired'. 698. cf. 漏 lòu ‘to leak’ (MC ləuC; LH loC; OC *rô(k)h). The unique vowel correspondence may indicate a relatively recent date of borrowing. 699. Siamese irregularly raised *o to /u/ before /-k/ in this etymon. The Saek form may be a Siamese borrowing. 702. Saek /hot4/ is a borrowing from Siamese or Lao. 704. cf. PAN *matáy ‘to die’ 709. cf. 塗 tú ‘to smear’ (MC duo; LH dɑ; OC *lâ). 711. Guigang, Long'an, and Wuming have /krapDS1/, /kʰlapDS1/, /klapDS1/ respectively. 713. Wuming /liawC2/ is a later borrowing from Mandarin. The Saek form /lɛ:wC2/ is a borrowing from Lao or Siamese, as suggested by the vowel /ɛ:/. 716. See '1s pronoun (strong)'. 717. PT had both *ku:A and *kawA. See Strecker (1984) for discussion. The distribution of the two may have been prosodically conditioned, i.e. *kawA occurred in prosodically prominent positions, while *ku:A occurred in prosodically weak positions. 718. See Strecker (1984) for discussion. This etymon probably occurred in prosodically prominent position, cf. the reduced form *maɰA. 720. See Strecker (1984) for discussion. This etymon probably occurred in prosodically prominent position, cf. the strong form *mɯnA. 722. The Saek reflex /o:/ may have resulted from a reduction of ealier *-aw. 723. This form might not be reconstructible at the PT level. 724. cf. 雙 shuāng ‘pair’ (MC ṣåŋ; LH ṣɔŋ; OC *srôŋ). 375 725. cf. 三 sām ‘three’ (MC sâm; LH sɑm/ səm; OC *sə̂ m). 726. cf. 四 sì ‘four’ (MC siC; LH sis; OC *sis/ slis). 727. Cf. 五 wǔ ‘five’ (MC ŋuoB; LH ŋɑB; OC *ŋâɁ). 728. cf. 六 liù ‘six’ (MC ljuk; LH liuk; OC *ruk; Baxter’s OC *C-rjuk). 729. cf. 七 qī ‘seven’ (MC tsʰjet; LH tsʰit ; OC *tsʰit < *snhit). 730. cf. 八 bā ‘five’ (MC păt; LH pɛt; OC *prêt). 731. cf. 九 jiǔ ‘nine’ (MC kjəuB; LH kuB; OC *kuɁ/ kwəɁ). 732. cf. 幾 jǐ ‘how many’ (MC kjeiB; LH kɨiB; OC *kəjɁ). The tonal reflex is rather irregular because this is a grammatical item. 736. The vowel reflexes are not completely regular because this etymon is a grammatical word. 737. cf. 多 duō ‘many’ (Schuessler 2009) from MC tâ, LH tɑ < tɑi, and OC *tâi < tlai. 738. See 'small'. 740. Wuming has /klaŋB2/ 'half' 743. The Saek form is a borrowing from Lao/Siamese. 747. cf. 線 xiàn ‘thread’ (MC sjänC; LH sianC; OC *sans). 750. This etyma is attested widely in NT, e.g. /pajA2/ in Nandan, Hengxian, Guigang, etc. 751. cf. 正 zhēng ‘five’ (MC tśjäŋ; LH tśeŋ; OC *teŋ). This might be in fact be a post-PT borrowing (Manomaivibool 1976). 753. Many dialects in Northwestern Vietnam unexpectedly have /-e-/, e.g. Lao, Black Tai, White Tai, and Sapa. 756. cf. 早 zǎo ‘morning’ (MC tsâuB; LH tsouB; OC *tsûɁ). 759. In NT, this etymon has been generalized and now mens that 'cooked rice'. 763. The final/-n/ in Siamese is unetymological and was acquired under the influence of the deictic /ni:C2/ ‘this’ in the expression /mɯəB2 wa:nA2 ni:C2/ ‘this past yesterday’. Also see Li (1977: 128-129). 376 764. The final /-ŋ/ in Siamese is aberrant and must have been due to the influence of /ni:C2/, cf. the expression /wanA2 phruŋB2 ni:C2/. 765. See 'yesterday'. 766. Wuming /pi:A1/ shows an unexpected vowel reflex (see §5.4). 770. SWT dialects show reflexes of *n-. This is an innovation due to the influence of *l.no:k 'outside', which is reflected as *nɔ:kD in PSWT. 771. This cannot go back to *C.n- because Saek has /l-/ rather than /n-/. 774. SWT and some CT dialects show reflexes *i:. This is probably because this etymon is a grammatical word. It is also possible that two forms *najC and *ni:C co-existed. See *kawA and *ku:A both '1s pronoun'. 775. Some languages unexpectedly have /ɤ/, e.g. Sapa. 780. Many dialects such as Lao, Yuan, Western Nung have /bo:B1/ rather than /bawB1/. These monophthongal reflexes are reduced forms developed independently in individual varieties. Lungchow and Yay have both the full and reduced forms. See 'not (strong 2)’. 781. See 'not strong (1)'. 782. The tonal irregularity is not surprising because this etymon is a grammatical item. 786. Wuming has /kɤC1/. Some dialects have irregular tonal reflexes because this etymon is a grammatical item. 377 378 APPENDIX C INDEX OF GLOSSARY The numbers in the rightmost column are the identifying numbers in Appendix B. acacia alive all also announce, to answer, to ant antelope ape arm arrive, to ascend to ashes (wood) ashes, leaf ashgourd ask, to asleep *ge:A *ɲaŋA *daŋA *ko:C *pa:wB *χa:nA *mɤcD *wa:lA *kaŋA *qe:nA *C̬ .tɤŋA *χɯnC *dawB *brawB *wakD *c.ra:mA *ɗakD 186 494 736 786 517 518 123 84 86 25 584 586 214 213 181 519 476 astringent in taste awaken (someone), to axe back bad bag bamboo bamboo hat bamboo shoot bamboo strip for tying or weaving banana banana blossom bank, river banyan bark, to basin bathe, to *ʰwɯətD *plokD *xwa:nA *ʰlaŋA *rwɤ:jC *croŋA *prajB *klɤpD *r.na:ŋA *b.twu:kD *kluəjC *pli:A *ʰwaŋB *rajA *ʰrawB *ʔa:ŋB *ʔa:pD 435 699 328 37 448 342 179 359 170 355 177 158 239 187 526 338 693 be (in a place), to be finished, to be left over, to be, to; b̩ ecome, to beach, sandbar bear beard bedbug bee before beg, to believe, to below bend over, to big (1) big (2) bird *ˀju:B *le:wC *ʰlɯəA *m̩ .pelA *ha:tD *ʰmwɯjA *momB *C̬ .rɯətD *to:B *ko:nB *k.ro:A *ɟɯǝB *lwɤ:ŋB *kɤmC *ʰluəŋA *ɓɯ:kD *C̬ .nokD 625 713 697 624 240 80 21 117 121 775 520 621 769 564 406 407 97 379 bite (n.), speech bite, to bitter black blind blistered blood blow, to board boat boil (intr.), to boil (tr.), to bone borrow, to brain bran break, to bright, light (n.) broom bruised (1) bruised (2) *gamA *C̬ .qɤpD *C̬ .qɤmA *C̥ .damA *ɓo:tD *bo:ŋA *lɯətD *powB *pe:nC *C̬ .rwɯəA *ɗɤ:tD *tomC *C̥ .dukD *ˀjɯ:mA *ʔe:kD *ramA *t.rakD *roŋB *ɲu:A *ɟamC *wokD 744 538 434 392 431 473 56 542 321 385 668 667 52 679 64 377 639 401 346 491 492 build, to burn (intr.), to burn (tr.), to burst, to butterfly buttock buy, to cane for walking carry (a bag), to carry in the arms, to carry on the back, to carrying pole (1) carrying pole (2) castrate, to caterpillar catfish cave centipede chase, to cheek chest *ko:B *ʰmajC *prawA *p.re:kD *ɓɤ:j *konC *z.ɟɯ:C *dawC *tri:wC *ʔu:mC *trwa:mA *tra:pD *ɢa:nA *to:nA *C̥ .boŋC *C̥ .dokD *cramC *q.sipD *lajB *ke:mC *ʔɤkD 547 664 665 640 138 43 677 343 561 560 562 334 335 634 136 103 231 139 712 17 36 chew, to chicken child (offspring) child (young person) chin, jaw chisel chop, to chopping board civet cat classifier for long, thin, object classifier for things classifier for tools clear, clean climb, to close (the eyes), to close, to cloud clump (as of bamboo) coarse, tough cockroach cockscomb *giəwC *kajB *lɯ:kD *ɗekD *ɢa:ŋA *siəwB *tramC *χiəŋA *ʰɲelA *selC *ʔalA *mwa:kD *saɰA *pi:nA *ʰlapD *hapD *ʰwɯəC *ko:A *ʰɲa:pD *sa:pD *ho:nA 532 93 270 284 20 329 637 325 90 747 745 748 399 601 548 610 247 168 421 140 150 380 cockspur coconut grub cold (n.) comb come into contact, to come, to cook (rice), to cooked in bamboo tube cord, string core cough, to count, to cover (with cloth), to crab crawl over, to crawl, to crazy creek crocodile crooked cross over, to *t.nɤ:jA *ɗuəŋC *q.watD *ʰrwɯ:jA *C̬ .tɯ:kD *ʰma:A *truŋA *ʰla:mA *sa:jA *ke:lB *ʔajA *napD *hɤmB *pɯwA *twajB *g.lwɤ:nA *ɓa:C *ro:ŋB *ŋɯəkD *gotD2 *kwa:B 151 134 72 344 650 582 670 384 746 439 544 514 675 105 600 599 486 237 114 428 591 cross, to crossbow crow crow, to cucumber, melon cut, to dam dark dark (red) daughter-in-law day day after tomorrow day before afterday daytime deaf debt deceive, to deep deer defeat, to; be defeated, to deficient *χa:mC *ʰnwɯəC *ka:A *χalA *p.rwe:ŋA *tacD *hwa:jA *mɯ:tD *klamB *baɰC *mwɯ:C *C̬ .rɯ:A *zɯ:nA *ŋwanA *ʰnuəkD *ʰni:C *bra:ŋA *lɤkD *kwɯəŋA *be:C 590 327 98 527 180 635 368 400 396 282 755 765 762 753 433 304 676 422 83 680 *bro:ŋB 446 descend, to dew, mist die, to differ, to dike between rice fields disappear, to disgusted ditch dive, to divide, to do not dog door downsteam, below dream, to dried up drive away, to drop, to drum drunk (1) drunk (2) *n̩ .loŋA *C̬ .nwa:jA *p.ta:jA *ta:ŋB *ɣalA *ʰrwɤ:jA *ɓɯəB *ʰmɯəŋA *ɗamA *panA *ʰɲa:B *ʰma:A *tu:A *taɰC *hwanA *ʰre:ŋC *k.rapD *tokD *klo:ŋA *mawA *mwi:A 587 257 704 513 367 703 485 366 609 632 784 75 313 768 620 463 711 592 390 469 470 381 dry dry land dry season duck duckweed dumb dust off, to dye, to each other, together ear ear (of rice) earth earthworm easy eat, to eaves edge egg eight elbow elephant *χaɰB *ɓokD *C̬ .le:ŋC *pitD *ʰne:A *ɓɤɰC *pacD *ɲwu:mC *kanA *krwɯ:A *rwɯ:ŋA *ɗin *t.nɯəlA *ŋa:jB *kɯɲA *ʑa:jA *ri:mA *qraj *pe:tD *C̬ .swo:kD *ɟa:ŋC 464 224 262 95 204 467 687 497 741 18 171 221 126 468 531 312 778 148 730 26 82 embroider, to empty entangled enter, to entrust, to epidemic erect, to excrement exit, to expensive expose to the sun, to extinguish, to eye face face down face up fall down, to fall off, to familiar fang far *se:wB *plɤwB *ɲuŋB/C *χawC *ʰwa:kD *ra:B *taŋC *C̬ .qɯjC *ʔo:kD *be:ŋA *p.ra:kD *ɗapD *p.ta:A *ʰna:C *qwamC *ʰŋa:jA *lɤmC *ʰlɤnB *gunC *χe:wC *k.lajA 499 444 472 588 715 73 657 50 589 474 706 673 11 10 454 455 594 593 475 19 453 fart fat father, man father's younger brother father's younger sister fear, to (1) fear, to (2) fern field, dry field, open field, paddy finger, toe fingernail, toenail fire firewood, bamboo firewood, hard fish fish hook fish net fishbone fishtrap *k.tɤtD *bwi:A *bo:B *ʔa:wA *ʔa:A *ʰla:wA *ja:nC *ku:tD *rɤjB *doŋB *na:A *ni:wC *C̬ .lepD *wɤjA *ʰlɯəwA *wɯ:lA *pla:A *ɓetD *kre:A *ka:ŋC *zajA 51 403 263 278 276 626 627 203 363 364 362 28 29 207 212 211 102 379 380 147 381 382 fishy five flame flea float, to flood, to flour flow, to flower fly, to flying squirrel fog fold double, to fold, to foot forehead forest forget, to four fragrant frog *ɣa:wA *ha:C *ple:wA *ʰmatD *wu:A *C̬ .tuǝmB *ɓɯəA *ʰlwajA *ɓlo:kD *ɓilA *ba:ŋB *ʰmo:kD *dopD *bapD *ti:nA *pra:kD *ɗoŋA *lɯ:mA *si:B *ho:mA *kɤpD 443 727 208 130 607 603 308 602 157 598 87 248 655 654 34 7 226 619 726 440 108 frog, small frog, tree fruit full gadfly galangal gall bladder, bile garden ginger girl give, to gizzard go hungry, to go upward, to go, to goiter good goose gradually, slowly grain granary *krwe:C *pa:tD *ʰma:kD *k.temA *ʰlɯəkD *xa:B *ɓli:A *swɯ:nA *χi:ŋA *ɓɯ:kD *haɰC *p.taɰA *ʔɤtD *mɯəA *pajA *ʰniəŋA *ɗɤjA *ha:nB *gɔ:jB *m.lecD *ˀjɯǝwC 109 110 163 445 127 194 70 365 193 288 686 149 612 585 583 23 447 96 779 165 310 grandfater, maternal grandfather, paternal grandmother, maternal grasp, to; perch, to grass grass, thatch grasshopper gray gray-haired grease, fat great-grandchild great-grandparent green grove, wood grow, to; rise, to gum; gill gust (of rain) gutter, trough hail hair, body; feather hair, head *ta:A *pɯwB *na:jA *capD *ʰɲɯəC *ɣa:A *p.takD *ʰmo:ŋA *ʰŋwu:kD *manA *ʰlenC *ɟɯəC *xiəwA *pa:B *ʰma:C *ʰŋɯəkD *kra:B *rwɯəŋA *trepD *q.pɯlA *prɤmA 268 267 269 555 200 201 118 397 6 219 273 271 394 227 701 15 253 383 258 5 3 383 hair, pubic half hammer, to hand handle (of a knife) handle, rod hang down, to hang up, to hard hate, to have, to hawk head (1) head (2) heap up, to hear, to heart, breath heavy heel hemp hen, young *ʰmwu:jA *grɤŋB *ɣo:lC *mwɯ:A *ɗa:mC *galA *ho:jC *qwe:nA *k.re:ŋA *ɟaŋA *mi:A *lamB *krawC *truəA *ko:ŋA *ŋinA *cɤɰA *ʰnakD *sɤnC *pa:nB *C̬ .qɤ:ŋB 44 740 644 27 332 333 708 707 420 613 681 101 1 2 515 616 65 410 35 349 94 hoarfrost hold in the closed hand, to hold in the jaws, to hold in the mouth, to hold the breath, to hole hole, crack hook horn horse hot house how many, several hug, to; embrace, to hump of an ox hundred; string, to hungry hunt, to hurt (1) hurt (2) husk *ʰmɯǝjA *kamA *ga:pD *ʔɤmA *klanC *ru:A *ɟo:ŋB *xɔ:A *qawA *ma:C *rwu:lC *rɤ:nA *ki:C *ko:tD *ʰno:kD *ro:jC *ˀjɯǝkD *p.rawB *cepD *ke:tD *ka:pD 256 552 551 537 545 232 233 330 141 79 429 309 735 546 152 732 477 506 487 488 161 idle, free illness, fever impetigo imprison, to incubate, to indigo (1) indigo (2) insect inside insipid intestine iron (1) iron (2) itch, to (1) itch, to (2) jackal join, to joint jug kidney-stones kill, to *ɗwɤ:jA *k.raj *kritD *k.raŋA *wakD *g.ra:mA *kromC *m.le:ŋA *C̥ .daɰA *cɯ:tD *sajC *ʰlekD *mwa:A *ɣalA *ɣomA *nwajA *to:B *χo:C *krajA *ʰni:wC *qa:C 484 71 61 505 550 353 354 122 770 438 67 215 216 628 629 81 631 30 339 74 705 384 kiss, to knee knife knot, hair know, to lacquer lady last (year) late late at night laugh, to lay (a cloth, etc.) across, to lazy leaf leaf for wrapping (big) leak, to lean back, to leech, aquatic leech, land leg lick, to *cu:pD *χowB *mitD *klawC *rɯ:wC *rakD *na:ŋA *kla:jA *ʰla:C *ɗɯkD *krɯəwA *ba:tD *kli:kD *ɓaɰA *k.to:ŋA *rwo:B *ʔi:ŋA *pli:ŋA *da:kD *p.qa:A *C̬ .lwiəA 540 32 323 4 623 358 289 752 451 758 528 710 482 155 156 698 566 116 115 31 533 lie down, to, sleep, to lift, to light (in weight) lightning liquid, soft liquor litter (of young) little, few liver long long (of time) loom loose lost louse, body louse, chicken louse, head low, short (not tall) lung lye machete, big knife *nwɯ:nA *ɲo:A *C̥ .bawA *m.le:pD *ʰle:wA *ʰlawC *gro:kD *ʰno:jB *tapD *rɯjA *hɤŋA *trukD *ʰlo:mA *ʰloŋA *m.lelA *rwɤjA *trawA *tamB *pwɤtD *ɗaŋB *ɟm̩ .ra:C 549 559 411 254 460 305 749 738 68 404 449 350 423 493 131 132 128 409 66 218 324 make a fire, to make noise, to man, male many, much mark, to; aim, to marking, patterns marrow, bone master, owner matter meal, early meal, evening meal, late measure from thumb to fingertip meat, flesh medicine middle millet moan, to monkey moon, month moonlight *ɗaŋA *ɗaŋA *ʑa:jA *ʰla:jA *ʰma:jA *C̬ .la:jA *ʔwu:kD *ɕɤwC *ɣwa:mA *ŋa:jA *ɟm̩ .rawA *C̬ .lwi:ŋA *ɣɯ:pD *n.mɤ:C *ˀjɯəA *kla:ŋA *ʰwɯǝŋC *gra:ŋA *li:ŋA *ɓlɯənA *ʰŋa:jA 672 512 285 737 507 361 53 294 788 759 761 760 743 57 306 772 175 525 85 242 245 385 morning glory morning, early mortar (1) mortar (2) mosquito moss, aquatic mother; woman mother's younger sibling mountain mountain, stone mouse, rat mouth mucus of the nose mud mushroom mustar green mute name name of first month navel (1) navel (2) *ɓuŋC *ɟawC *grokD *ʰrumA *ɲuŋA *dawA *me:B *na:C *C̥ .do:jA *pra:A *ʰnu:A *pa:kD *mu:kD *lɤmB *ʰrwetD *ka:tD *ŋwamC *ɟɤ:B *ciǝŋA *ɗwɯ:A *ɓli:C 189 756 373 374 119 205 264 277 229 230 77 12 9 241 202 190 466 293 751 41 42 near neck needle nephew, niece, or grandchild new nibble, to; peck, to night night night blind nine nose not (strong 1) not (strong 2) not (weak) not yet obtain, to old (of living beings) old (of things) one one or so, any open (the mouth), to *k.raɰC 452 order, to; blow the nose, *saŋB 521 *ɣo:A 22 to orphan *gm̩ .ra:C 292 *qemA 356 other *ʔɯ:nB 742 *ʰla:nA 272 otter *na:kD 91 *ʰmɤ:lB 414 outside *l̩ .no:kD 771 *tɔ:tD 539 owl *gawC 100 *ɣɯ:nA 754 pair *gu:B 734 *ɣamB 757 pangolin *lilB 92 *ʰwa:ŋA 432 parent's older brother *luŋA 274 *kɤwC 731 parent's older sister *pa:C 275 *ɗaŋA 8 partition, lid *hwa:A 316 *ɓawB 780 peel, bark *plɯəkD 160 *bo:B 781 peel, to *po:kD 643 *mi:A 782 penis *ɣwajA 46 *pajB 783 person, human being *ɢwɯnA 283 *ɗajC 683 pestle *sa:kD 375 *ke:B 415 pick up, to *ˀjipD 554 *kawB 416 pick, to *kepD 684 *nɯ:ŋB 723 pickle, to *ɗo:ŋA 660 *sakD 739 pig *ʰmu:A 76 *ʔa:C 530 pillar *sawA 315 386 pinch, to pit place, ground plant, to pliers, thongs plough pluck, to plug (a hole), to plugged plum point, to poison (fish), to pomelo porcupine pot pound (rice), to pound, to pound, to (1) pound, to (2) pour, to pronoun, 1p *ʰni:pD *C̬ .kumA *di:B *plwɯ:kD *gi:mA *crwajA *C̥ .decD *ʔwɯtD *sakA *manC *ɟi:C *C̥ .bɯəA *bu:kD *ʰmenC *ʰmo:C *zo:mC *tamA *to:kD *dupA *rwa:tD *rawA 557 234 228 503 331 370 641 658 490 184 567 509 185 89 340 646 645 649 648 605 722 pronoun, 1s (strong) pronoun, 1s (weak) pronoun, 2s (strong) pronoun, 2s (weak) pronoun, 3s (strong) pronoun, 3s (weak) pubic mound pull, to; drag, to puppet, marionette pus put out of the mouth, to put, to putrid quit, to raft rain raise, to raise, to raisin ralated by marriage rattan *kawA *ku:A *maɰA *mɯŋA *manA *mɯnA *ʰnawB *C̬ .la:kD *hunB *ʰno:ŋA *ɣa:jA *s.cɤ:lB *xwi:wA *ˀja:B *be:A *C̥ .wɯnA *liəŋC *ɲɤkD *ʔitD *t.no:ŋA *C̥ .wa:jA 717 716 719 718 721 720 48 563 360 60 534 570 442 614 386 252 511 558 197 290 198 raw, not ripe recite, to recognize, to red reed repeat, to resin rest, to revolve, to; spin, to rice rice, husked rinse, to river road roast, to roof, to root rope, cord rotten saddle saliva *C̥ .dipD *ʔa:nB *cakD *C̥ .dwi:ŋA *ʔo:C *zamC *ˀja:ŋA *bakD *panB *C̬ .qawC *sa:lA *C.lwɤ:ŋC *da:B *ʰrwɤnA *ˀjɯəŋC *mwuŋA *C̬ .ra:kD *ɟɤ:k *nawB *ʔa:nA *la:jA 419 529 622 391 199 575 176 695 611 172 174 690 238 387 666 674 167 336 461 389 16 387 salt salt, to sand satiated scald, to scales, fish scold, to; revile, to scoop, to scratch, to scurf see, to seedling, rice seek, to sell, to senior separate, to sesame seven sew, to shade shadow, reflection *klwɯəA *ʔɯǝpD *zwɯəjA *ʔi:mB *luəkD *klecC *ɗa:B *takD *kawA *ɣɯjA *tranA *kla:C *kraA *p.qa:j *crawC *bra:kD *r.ŋa:A *cetD *ɲepD *rɤmB *ŋawA 217 659 222 478 669 146 522 662 569 63 615 173 630 678 417 633 195 729 498 260 261 shake, to shaman sharp, sharp edge sharp-pointed sheath, pod shellfish shin, lower leg shoot, to short (not long) shoulder shrimp shrimp, small shrink, to shuttle of loom sibling, older sibling, younger side side, ribs sift, to silkworm single, only one *C̥ .wajA *ʰmo:A *ɣɤmA *ʰle:mA *q.wakD *ho:jA *ɣe:ŋB *ɲɯŋA *tinB *C̥ .ba:B *kuŋB/C *ŋiəwA *ʰrotD *p.rawB *bi:B *nwo:ŋC *C̥ .bɯǝŋC *k.ra:ŋC *qrɤŋA *mo:nC *ɗiəwA 577 298 457 456 164 104 33 508 405 24 106 107 702 351 265 266 773 55 504 137 733 sink, to sit, to six skin sky, heaven sky, weather slap, to slash, to slave (1) slave (2) slip and fall, to slippery, smooth slow small smear, to smell (tr.), to smoke snake snap, to snare snore, to *cɤmA *naŋB *krokD *ʰnaŋA *ɓɯnA *va:C *tɤpD *walA *χɔ:jB *kra:C *bla:tD *m.lɯ:lB *na:nA *no:jC *da:A *ɗɤmA *ɣwanA *ŋwɯ:A *ɓli:tD *re:wC *klɤlA 608 581 728 58 250 251 647 636 295 296 595 458 450 408 709 618 209 113 571 382 543 388 soak, to (1) soak, to (2) soil son-in-law soot soul; whorl in the hair sour sow, to; scatter, to spear spider spill, to spirit (1) spirit (2) spleen split bamboo flooring spool sprout, shoot squeeze, to squirrel stairs, ladder stake *ɟe:B *ʰma:B *tɯmA *C̬ .kɯəjA *ʰmi:C *qwanA *sɤmC *C̥ .wa:lB *kro:kD *krwa:wA *ɓa:B *pri:A *mwa:ŋA *ma:mC *wa:kD *ʰlwu:tD *ʰno:B *galC *ro:kD *ɗrwajA *ʰlakD 663 661 220 281 210 301 436 501 326 133 604 299 300 69 317 352 169 556 88 314 311 stamp (on), to stand, to star (general) star (in the sky) starfruit steal, to steam, to steam, vapor steep stem step on, to (1) step on, to (2) sterile sticky stiff and tired stinger (of a bee) stink bug stinky stomach, belly stone stool *dɯ:pD *C̥ .ɟɯ:nA *t.na:wA *ɗi:B *ɢwɯəŋA *C̬ .lakD *ʰnɯŋC *s.ʔwɤ:jA *ʰliŋB *ka:nC *ɲamB *ˀjiǝpD *ʰmanA *ʰniəwA *mɯəjB *lajA *ge:ŋ1 *ʰmenA *dwu:ŋC *tri:lA *taŋB 572 596 243 244 183 685 671 259 425 159 573 574 471 459 480 153 135 441 40 223 319 straight strand (of rope) straw, stubble stream, mountain stretch out, to stretch, to stroke, to; carress, to strong, strength stuck stump sugarcane sunshine swallow (n.) sweat sweep, to sweet, delicious swim, to swollen tail take a bite, to take down, to; put down, to *zɤ:B *kliəwA *wɯǝŋA *qrwɤjC *ˀjiətD *ɲɯ:tD *C̬ .lu:pD *re:ŋA *ga:A *to:A *ʔo:jC *C̥ .dwi:tD *ʔe:nB *r.tɯəB *kwa:tD *C̥ .wa:lA *lo:jA *gaɰB *trwɤ:ŋA *katD *ploŋA 427 357 378 236 580 656 568 495 426 166 178 246 99 59 688 437 606 479 145 536 642 389 take up in the two cupped hands, to take, to taro taste, to tea teach, to tend animal, to tendon termite testicles thick thin (not fat) thin (not thick) thing this thorn thread (1) thread (2) three thunder tick *ko:pD *ʔawA *prɯəkD *ɟimA *ɟa:A *so:lA *ɟɯǝŋC *ˀjenA *mo:tD *tram *ʰna:A *pro:mA *C̥ .ba:ŋA *χɔ:ŋA *najC *ʰna:mA *C̥ .da:jC *ʰmajA *sa:m *pra:C *trepD 553 682 192 617 196 694 510 54 124 47 412 402 413 322 774 162 347 348 725 255 129 tie up, to tie, to tight, narrow time (classifier) tip, end tip, highest point tired, bored tired, exhausted to commission tomorrow tongue tooth top for spinning torn tortoise, water township track transplant, to tree, wood tremble, to tripod *la:mB *cm̩ .rukD *gapD *bajA *pla:jA *ɲo:tD *ʰna:jB *ʰnɯəjB *ʑaɰC *ɟm̩ .rukD *li:nC *wanA *k.ra:ŋB *qa:tD *ʰwɯəA *mɯəŋA *rwu:jA *t.namA *mwajC *salB *giǝŋA 652 651 424 750 776 777 483 481 714 764 13 14 337 489 112 303 388 502 154 576 320 tube, bamboo turn upside down or inside out, to turtle tusk, ivory twist, to; wring, to two undo, to; untie, to unmarried man unmarried woman upstream, above urine vagina vegetable Vietnamese village vinegar vomit, to waist (1) waist (2) wait, to wake up, to *baŋ B/C *pli:nC 341 565 *tawB *ŋa:A *ɓitD *so:ŋA *ke:C *ɓa:wB *sa:wA *ʰnɯəA *niəwB *hi:A *prakD *ke:wA *ɓa:nC *ʰmi:B *rwɯəkD *ˀje:wA *C̥ .wɯǝtD *cra:C *k.tɯ:nB 111 142 579 724 653 286 287 767 49 45 188 297 302 307 535 38 39 696 700 390 walk, to *pra:jC warm *ʔunB warn,to *tɯǝnA wash (clothes), to *zakD wash, to *za:wA wasp *b.twi:lA water *C̬ .namC water buffalo *ɣwa:jA water pipe *linA weave, to (baskets, mats) *sa:n weave, to (cloth) *tamB weed, to *ɓla:jA weep, to *t.hajC weigh, to *ɟaŋB well, spring *ɓo:B wet *domA which *ɗaɰA white *xa:wA white-spotted *ɓla:ŋB widowed *ʰma:jC wife *miəA 597 430 523 691 692 120 206 78 369 500 496 638 524 516 235 465 785 393 398 291 280 wife of man's younger brother wild wild olive wind window wing (1) wing (2) winnowing basket wipe, to with, and withered worm wound wrestle, to writing, book yam yawn, to year yellow yesterday yoke *lɯǝwA *C̬ .tɯənB *ko:kD *C̬ .lɯmA *ta:ŋB *wɯətD *pi:kD *ɗoŋC *ɟetD *kapD *ʰriəwB *ʰno:lA *ɓa:tD *plamC *sɯ:A *manA *ʰra:wA *pi:A *ʰlɯəŋA *ŋwa:A *ʔe:kD 279 225 182 249 318 143 144 376 689 787 462 125 62 578 345 191 541 766 395 763 371 yoke, part of young, soft *ʔo:ŋC *ʔwu:nB 372 418 REFERENCES Abramson, Arthur .S. (1978). Static and dynamic acoustic cues in distinctive tones. Language and Speech, 21, 319-325. Abramson, Arthur S. (1962). The vowels and tones in Standard Thai: Acoustical measurements and experiments. International Journal of American Linguistics, 28(2), Part II. Abramson, Arthur S., & Erickson, Donna M. (1978). Diachronic tone splits and voicing shifts in Thai: Some perceptual data. Status report on speech research, Haskins Laboratories SR-53(2), 85-96. Aikhenvald, Alexandra Y., & Dixon, R. M. W. (2001). Introduction. In A. Y. Aikhenvald & R. M. W. Dixon (Eds.), Areal diffusion and genetic inheritance: problems in comparative linguistics (pp. 1-23). Oxford; New York: Oxford University Press. Alves, Mark J. (2007). A look at North-Central Vietnamese. In R. Wayland, J. Hartmann & P. Sidwell (Eds.), Papers from the 12th Annual Meeting of the Southeast Asian Linguistics Society 2002 (pp. 1-7). Canberra: Pacific Linguistics. Awơi-hathe, Aviong, A-Tý, A-Ly, Cobbey, Maxwell, & Cobbey, Vurnell. (1977). Northern Roglai Vocabulary = Suraq Vungã Sanãp Radlai = Ngữ-Vựng Rơglai. Manila, Philippines: Summer Institute of Linguistics. Ball, Douglas. (2003). Insights into the Patterns of Sounds: Stress in Tondano. Journal of Undergraduate Research, University of Rochester, 2, 15-19. Barrack, Charles M. (2002). The Glottalic Theory revisited: A negative appraisal. Indogermanische Forschungen, 107, 76–95. Baxter, William. (1992). A handbook of Old Chinese phonology. Berlin; New York: Mouton de Gruyter. Belotel-Grenié, Agnès, & Grenié, Michel (1994, September 18-22, 1994). Phonation types analysis in Standard Chinese. Paper presented at Third International Conference on Spoken Language Processing (ICSLP 94), Yokohama, Japan. Benedict, Paul K. (1942). Thai, Kadai and Indonesian: A new alignment in Southeastern Asia. American Anthropologist, 44, 576-601. 391 Benedict, Paul K. (1975). Austro-Thai language and culture, with a glossary of roots. New Haven: HRAF Press. Benedict, Paul K. (1982). Qiang monosyllabization: a third phase in the cycle. Linguistics of the Tibeto-Burman Area, 7(2), 113-114. Benedict, Paul K. (1988). Kadai linguistics: the rule of engagement. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: Linguistic studies beyond Tai (pp. 323-340). Dallas: Summer Institute of Linguistics and the University of Texas at Arlington Publications in Linguistics. Benedict, Paul K. (1997a). Linguistic prediction: The case of Saek. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: The Tai branch (pp. 161-189). Arlington, TX: Summer Institute of Linguistics and the University of Texas at Arlington. Benedict, Paul K. (1997b). Tai Languages: Key to Austro-Tai. In A. Abramson (Ed.), Southeast Asian Linguistics Studies in honor of Vichin Panupong (pp. 25-28). Bangkok: Chulalongkorn University Press. Bennett, J. Fraser. (1994). Iambicity in Thai. Studies in the linguistic sciences, 24(1), 39-57. Bennett, J. Fraser. (1995). Metrical foot structure in Thai and Kayah Li: Optimalitytheoretic studies in the prosody of two Southeast Asian languages. Unpublished Ph.D. dissertation, University of Illinois at Urbana-Champaign, Urbana-Champaign. Booij, Geert. (1997). The role of the prosodic word in phonotactic generalizations. In T. A. Hall & U. Kleinhenz (Eds.), Studies on the phonological words (pp. 4772). Amsterdam; Philadelphia: Johnn Benjamins Publishing Company. Boonsner, Thepbangon. (1984). The phonological description of Yooy. Unpublished M.A. thesis, Mahidol University, Nakhorn Pathom. Bradley, David. (1982). Register in Burmese. In Papers in Southeast Asian Linguistics No.8, Tonation (pp. 117-132). Canberra: Pacific Linguistics. Brown, J. Marvin (1965). From Ancient Thai to modern dialects. In From Ancient Thai to modern dialects, and other writings on historical Thai linguistics (pp. 69-254). Bangkok: White Lotus. 392 Brugman, Johanna C. (2009). Segments, tone and alignment in Khoekhoe prosody. Unpublished Ph.D. dissertation, Cornell University, Ithaca, NY. Brunelle, Marc. (2008). Monosyllabicization in Eastern Cham. In W. Khanittanan & P. Sidwell (Eds.), SEALS XIV, Vol I: Papers from the 14 th annual meeting of the Southeast Asian Linguistics Society 2004 (pp. 43-58). Canberra: Pacific Linguistics. Brunelle, Marc. (2009). Tone perception in Northern and Southern Vietnamese. Journal of Phonetics, 37(1), 79-96. Burenhult, Niclas. (2005). A grammar of Jahai. Canberra: Research School of Pacific and Asian Studies, Australian National University. Chamberlain, James R. (1975). A new look at the history and classification of the Tai dialects. In J. G. Harris & J. R. Chamberlain (Eds.), Studies in Tai linguistics in honor of William J. Gedney (pp. 49-60). Bangkok: Central Institute of English Language, Office of State Universities. Chamberlain, James R. (1984). The Tai dialects of Khammouane Province, Laos: Their diversity and origins. The Science of Language, 4, 62-95. Chamberlain, James R. (1991). The efficacy of the P/PH distinction for Tai languages. In J. R. Chamberlain (Ed.), The Ram Khamhaeng controversy (pp. 453-486). Bangkok: Siam Society. Chamberlain, James R. (2000). The origin of the Sek implications for Tai and Vietnamese history. In S. Burusphat (Ed.), Proceedings: the International Conference on Tai Studies, July 29-31, 1998. (pp. 97-127). Bangkok: Institute of Language and Culture for Rural Development, Mahidol University. Chang, Kun. (1975). The reconstruction of Proto-Miao-Yao tones. Bulletin of the Institute of History and Philology, Academia Sinica(44), 541-628. Chantanakomes, Veena. (1980). A description of Moken: A Malayo-Polynesian language. Unpublished M.A. thesis, Mahidol University, Nakhon Pathom. Chao, Yuen-Ren. (1930). A system of tone-letters. Le maître phonétique, 45, 24-27. Chinese Academy of Sciences. (1959). Bùyīyǔ diàochá bàogào [Research reports on Bouyei languages]. Běijīng: Science Press. 393 Cho, Young-mee Y., & King, Tracy H. (2003). Semisyllables and universal syllabification. In C. Féry & R. van de Vijver (Eds.), The syllable in Optimality Theory. Cambridge; New York: Cambridge University Press. Clements, G. N. (1990). The role of syllabification cycle in core syllabification. In J. Kingston & M. E. Beckman (Eds.), Papers in Laboratory Phonology I: Between the grammar and physics of speech (pp. 283-333). Cambridge; New York, Port Chester, Melbourne, Sydney: Cambridge University Press. Clements, G. N., & Osu, Sylvester. (2002). Explosives, implosives, and nonexplosives: the linguistic function of air pressure differences in stops. In C. Gussenhoven & N. Warner (Eds.), Laboratory phonology 7 (pp. 299-350). Berlin; New York: Mouton de Gruyter. Cohn, Abigail. (in progress). The phonological structure of Indonesian truncations and abbreviation. Department of Linguistics, Cornell University. Cohn, Abigail, & McCarthy, John. (1998). Alignment and parallelism in Indonesian phonology. Working Papers of the Cornell Phonetics Laboratory, 12, 53-137. Conrady, August. (1896). Eine indochinesische Causative-Denominativ-Bildung und ihr Zusammmenhang mit den Tonaccenten. Leipzig: Otto Harrassowitz. Cooper, Adam. (to appear). Similarity avoidance in the Proto-Indo-European root. Proceedings of the 32nd Annual Penn Linguistics Colloquium, Penn Working Papers in Linguistics, 15(1). Croft, William. (1990). Typology and universals. Cambridge; New York: Cambridge University Press. Cushing, Josiah Nelson. (1914). A Shan and English Dictionary (2nd ed.). Rangoon: American Baptist Mission Press. Davison, Deborah S. (1991). An acoustic study of so-called creaky voice in Tianjin Mandarin. UCLA Working Papers in Phonetics, 78, 50-77. de Lacouperie, Terrien. (1886). The languages of China before the Chinese: Researches on the languages spoken by the pre-Chinese races of China proper previously to the Chinese occupation. London: D. Nutt. 394 de Lacy, Paul. (2004). Maximal words and the Māori passive. In J. McCarthy (Ed.), Optimality Theory in Phonology: A reader (pp. 495-512). Malden, MA: Blackwell. Dench, Alan. (2001). Descent and diffusion: the complexity of the Pilbara situation. In A. Y. Aikhenvald & R. M. W. Dixon (Eds.), Areal diffusion and genetic inheritance: problems in comparative linguistics (pp. 105-132). Oxford; New York: Oxford University Press. Dhananjayananda, Puttachart. (1997). The emergence of the length distinction in midfront vowels e-ee in Thai. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: The Tai branch (pp. 327-336). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. Diffloth, Gérard. (1976). Minor-syllable vocalism in Senoic languages. In P. N. Jenner, L. C. Thompson & S. Starosta (Eds.), Austroasiatic Studies (pp. 229248). Honolulu: The University Press of Hawai'i. Diffloth, Gérard. (1977). Towards a history of Mon-Khmer: proto-Semai vowels. Tōnan Azia Kenkyū, 14(4), 463-495. Diffloth, Gérard. (1984). The Dvaravati Old Mon language and Nyah Kur. Bangkok: Chulalongkorn University Printing House. Diffloth, Gérard. (1991). Palaungic vowels in Mon-Khmer perspective. In J. H. C. S. Davidson (Ed.), Austroasiatic languages: Essays in honour of H. L. Shorto (pp. 13-27). London: School of Oriental and African Studies. Diller, Anthony. (2000). The Tai language family and the Comparative Method. In S. Burusphat (Ed.), Proceedings : the International Conference on Tai Studies, July 29-31, 1998. (pp. 1-32). Nakhon Pathom: Institute of Language and Culture for Rural Development, Mahidol University. Dixon, R. M. W, & Aikhenvald, Alexandra Y. (2002). Word: a typological framework. In R. M. W. Dixon & A. Y. Aikhenvald (Eds.), Word: A crosslinguistic typology (pp. 1-41). Cambridge: Cambridge University Press. Đoàn, Thiin Thuut. (1996). Tay-Nung language in the north Vietnam. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies. 395 Ðoàn, Thiện Thuật. (1992). Ngữ âm tiing Viit [Vietnamese phonology]. Hà NNi: Nhà xuất bản Dại học Quốc gia Hà Nội [Hanoi National University Press]. Downing, Laura J. (2006). Canonical forms in prosodic morphology. Oxford; New York: Oxford University Press. Durie, Mark. (1985). A grammar of Acehnese on the basis of a dialect of north Aceh. Dordrecht, Holland; Cinnaminson, USA: Foris Publications. Edmondson, Jerold A., & Yang, Quan. (1988). Word-initial preconsonants and the history of Kam-Sui resonant initials and tones. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: Linguistic studies beyond Tai (pp. 143166). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. Edmondson, Jerry A., & Solnit, David B. (1988). Introduction. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: Linguistic studies beyond Tai (pp. 117). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. Edmondson, Jerry A., & Solnit, David B. (1997). Introduction. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: The Tai Branch (pp. 1-26). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. Ferlus, Michel. (1975). Vietnamien et proto-Viet-Muong. Asie du Sud-Est et monde insulindien, 6(4), 21-55. Ferlus, Michel. (1989). Remarques sur le consonantisme des langues kadai. Paper presented at the 22nd International Conference on Sino-Tibetan Languages and Linguistics. University of Hawai'i, Manoa. October 5-8, 1989. Ferlus, Michel. (1990). Remarques sur le consonantisme de proto thai-yay (révision du proto-tai de Li Fangkuei). Paper circulated at the 23rd International Conference on Sino-Tibetan Languages and Linguistics, University of Texas at Arlington. Ferlus, Michel. (1992). Histoire abregée de l’évolution des consonnes initiales du vietnamien. Mon-Khmer Studies, 20, 111-127. Ferlus, Michel. (1996a). Evolution vers le monosyllabisme dans quelques languages de l'Asie du Sud-Est. Séance du 23 novembre 1996, la Société de Linguistique de Paris. 396 Ferlus, Michel. (1996b). Remarques sur le consonantisme du proto kam-sui. Cahiers de linguistique Asie orientale, 25(2), 235-278. Ferlus, Michel. (1998). Nhahuen-French-English lexicon (edited and annotated by Pascale Jacq & Paul Sidwell). München: Lincom Europa. Ferlus, Michel. (1999). Les dialectes et les écritures des Tai (Thai) de Nghệ An (Vietnam). Communication de la Treizièmes Journée de Linguistique d'Asie Orientale, Paris (EHESS-CNRS). 10-11 juin 1999. . Ferlus, Michel. (2008). The dialects of Nghệ An, Vietnam (Tay Daeng, Tay Yo, and Tai Muong). In A. Diller, J. A. Edmondson & Y. Luo (Eds.), The Tai-Kadai languages. London; New York: Routledge. Fine Arts Department. (1986). Charuek Nai Prathet Thai [Inscriptions in Thailand] (Vol. 5). Bangkok: National Library. Flemming, Edward. (2004). Contrast and perceptual distinctiveness. In B. Hayes, R. Kirchner & D. Steriade (Eds.), Phonetically-based Phonology (pp. 232-276). Cambridge ; New York: Cambridge University Press. Freiberger, Nancy. (1964). A phonemic description of Nong (Nung). Te Reo, 7, 15-22. Freiberger, Nancy, & Bé, Vy Thị. (1976). Sẹc mạhn slứ nohng fạn slihng = Ngữ-vưng nùng phan slinh = Nung fan slihng vocabulary: Nùng-Viit-English Summer Institute of Linguistics: Viin Chuyên-Khảo Ngữ-Học [Institute of Linguistics]. Gamkrelidze, Thomas V., & Ivanov, Vjacheslav V. . (1972). Lingvističeskaja tipologija i rekonstrukcija sistemy indoevropejskix smyčnyx [Linguistic Typology and the Reconstruction of the Indo-European Occlusives]. In Konferencija po sravnitel’no-istoričeskoj grammatike indoevropejskix jazykov: Predvaritel’nye materialy, 15-18 [Working Papers of the Conference on the Comparative-Historical Grammar of the Indo-European Languages (12—14 December 1972)] (Vol. 15-18). Moscow: Nauka. Gandour, J., Potisuk, S., & Dechongkit, S. (1994). Tonal coarticulation in Thai. Journal of Phonetics, 22, 477-492. Gandour, Jack. (1977). On the interaction between tone and vowel length: Evidence from Thai dialects. Phonetica, 34, 54-65. 397 Gedney, William J. (1965). Indic loanwords in spoken Thai. Unpublished Ph.D. dissertation, Yale University, New Haven. Gedney, William J. (1972a). A checklist for determining tones in Tai dialects. In S. M. Estellie (Ed.), Studies in Linguistics in honor of George L. Trager (pp. 423438). The Hague: Mouton. Gedney, William J. (1972b). A puzzle in Comparative Tai Phonology. In J. G. Harris & R. B. Noss (Eds.), Tai phonetics and phonology. (pp. 52-57). Bangkok: Central Institute of English Language, Office of State Universities. Gedney, William J. (1989a). Evidence for another series of voiced initials in ProtoTai. In R. J. Bickner, J. Hartmann, T. J. Hudak & P. Peyasantiwong (Eds.), Selected papers on Comparative Tai Studies. Ann Arbor: Center for Southeast Asian Studies, the University of Michigan. Gedney, William J. (1989b). Future directions in Comparative Tai Linguistics. In R. J. Bickner, J. Hartmann, T. J. Hudak & P. Peyasantiwong (Eds.), Selected papers on Comparative Tai Studies (pp. 7-116). Ann Arbor: Center for Southeast Asian Studies, the University of Michigan. Gedney, William J. (1989c). The Saek language of Nakhon Phanom province. In R. J. Bickner, J. Hartmann, T. J. Hudak & P. Peyasantiwong (Eds.), Selected papers on Comparative Tai Studies (pp. 373-400). Ann Arbor: Center for Southeast Asian Studies, the University of Michigan. . Gedney, William J. (1989d). Speculations on early Tai tones. In R. J. Bickner, J. Hartmann, T. J. Hudak & P. Peyasantiwong (Eds.), Selected papers on Comparative Tai Studies (pp. 207-228). Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Gedney, William J. (1991). Comments on linguistic arguments relating to Inscription One. In J. R. Chamberlain (Ed.), The Ram Khamhaeng controversy (pp. 193226). Bangkok: Siam Society. Gedney, William J. (1993). Saek final -l: archaism or innovation? In T. J. Hudak (Ed.), William J. Gedney's Saek language: Glossaries, Texts, and Translations (pp. 917-974). Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Gedney, William J. (1995). Linguistic diversity among Tai dialects of southern Guangxi. In T. J. Hudak (Ed.), William J. Gedney’s Central Tai dialects : 398 glossaries, texts, and translations (pp. 803-822). Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Gedney, William J. (n.d.). A comparative Tai wordlist. manuscript. Goldsmith, John A. (1995). Phonological Theory. In J. A. Goldsmith (Ed.), The handbook of phonological theory (pp. 1-23). Cambridge, MA: Blackwell. Gordon, Matthews. (2001). A typology of contour tone restriction. Studies in Language, 25(3), 423-462. Gordon, Matthews, & Ladefoged, Peter. (2001). Phonation types: a cross-linguistic overview. Journal of Phonetics, 29, 383-406. Green, Anthony Dubach. (2005). Word, foot, and syllable structure in Burmese. In J. Watkins (Ed.), Studies in Burmese linguistics (pp. 1–25). Canberra: Canberra: Pacific Linguistics. Grierson, G. A. (1903). Linguistic survey of India. Calcutta: Office of the Superintendent of Government Printing House. Guion, Susan G., & Wayland, Ratree. (2004). Aerodynamic coarticulation in sound change or how onset trills can condition a falling tone. In A. Agwuele, W. Warren & S.-H. Park (Eds.), Proceedings of the 2003 Texas Linguistics Society Conference. Somerville, MA: Cascadilla Proceedings Project. Gussmann, Edmond. (1992). Resyllabification and delinking: The case of Polish voicing. Linguistic Inquiry, 23(1), 29-56. Hall, T. Alan. (1997). Phonotactics and the prosodic structure of German. In T. A. Hall & U. Kleinhenz (Eds.), Studies on the phonological words (pp. 47-72). Amsterdam; Philadelphia: Johnn Benjamins Publishing Company. Halle, Moris, & Vergnaud, Jean-Roger.1980. (1980). Three Dimensional Phonology. Journal of Linguistic Research, 1(1), 83-105. Handel, Zev. (1998). The medial systems of Old Chinese and Proto-Sino-Tibetan. Unpublished Ph.D. dissertation, University of California at Berkeley, Berkeley. Handel, Zev. (2003). Northern Min tone values and the reconstruction of "soften initials". Language and Linguistics, 4, 47-84. 399 Harris, James W. (1983). Syllable structure and stress in Spanish: a nonlinear analysis. Cambridge, MA: MIT Press. Harris, Jimmy G. (1975). A comparative word list of three Tai Nüa dialects. In J. G. Harris & J. R. Chamberlain (Eds.), Studies in Tai linguistics in honor of William J. Gedney (pp. 202-230). Bangkok: Central Institute of the English Language and Office of State Universities. Haudricourt, André-Georges. (1948). Les phonèmes et le vocabulaire du thai commun. Journal Asiatique, 236, 197-238. Haudricourt, André-Georges. (1952). Les occlusives velaires en thai. Belletin de la Société de Linguistique de Paris, 48, 86-89. Haudricourt, André-Georges. (1954a). Comment reconstruire le chinois archaïque. Word, 10, 351-364. Haudricourt, André-Georges. (1954b). De l'origine des tons en vietnamien. Journal Asiatique, 242, 69-82. Haudricourt, André-Georges. (1956). De la restitution des intitiales dans les langues monosyllabiques: le problème du thai commun. Bulletin de la Société de Linguistique de Paris, 52, 307-322. Haudricourt, André-Georges. (1958). Les Sek de la province du Cammon (Laos), migration thai ou déportation chinoise? Journal Asiatique, 246, 107-108. Haudricourt, André-Georges. (1960). Notes sur les dialects des région de Moncay. Bulletin de l'Ecôle Française d'Extrême-Orient, 50, 167-177. Haudricourt, André-Georges. (1961). Bipartition et tripartition des systèmes de tons dans quelques langages d'extrême orient. Bulletin de la Société de Linguistique de Paris, 56, 163-180. Haudricourt, André-Georges. (1963a). La langue Sek. In Proceedings of the 25th Congress of Orientalists (Vol. 4, pp. 414-415). Haudricourt, André-Georges. (1963b). Remarques sur les initiales complexes de la language Sek. Belletin de la Société de Linguistique de Paris, 58(1), 156-163. Haudricourt, André-Georges. (1966-1974). Daïque (daic). In R. Shafer (Ed.), Introduction to Sino-Tibetan (pp. 453-525). Wiesbaden: O. Harrassowitz. 400 Haudricourt, André-Georges. (1975). A propos du puzzle de W.J. Gedney. In J. G. Harris & J. R. Chamberlain (Eds.), Studies in Tai linguistics in honor of William J. Gedney (Vol. 252-258). Bangkok: Central Institute of English Language. Hayes, Bruce. (1995). Metrical stress theory : principles and case studies. Chicago: University of Chicago Press. Henderson, Eugénie J.A. (1952). The main features of Cambodian pronunciation. Bulletin of the School of Oriental and African Studies, University of London, 14, 149-174. Hoang, Van Ma. (1997). The sound system of the Tày language of Cao Bằng Province , Vietnam. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: The Tai branch (pp. 221-235). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. Hock, Hans Henrich. (1991). Principles of Historical Linguistics. Berlin; New York: Mouton de Gruyter. Hombert, Jean-Marie. (1977). A model of tonal systems. UCLA Working Papers in Phonetics, 36, 20-32. Hombert, Jean-Marie, Ohala, John J., & Ewan, William G. (1979). Phonetic explanations for the development of tones. Language, 55(1), 37-58. Hooper, Joan B. (1976). An introduction to Natural Generative Phonology. New York: Academic Press. Hualde, José Ignacio. (1991). On Spanish syllabification. In H. Campos & F. Martínez-Gil (Eds.), Current studies in Spanish linguistics (pp. 475-494). Washington, D.C.: Georgetown University Press. Hudak, Thomas J. (Ed.). (1991a). William J. Gedney’s the Tai dialect of Lungming : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Hudak, Thomas J. (Ed.). (1991b). William J. Gedney’s the Yay language : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. 401 Hudak, Thomas J. (Ed.). (1993). William J. Gedney’s the Saek languge : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Hudak, Thomas J. (Ed.). (1994). William J. Gedney’s Southwestern Tai dialects : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Hudak, Thomas J. (Ed.). (1995). William J. Gedney’s Central Tai dialects : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Hudak, Thomas J. (Ed.). (1996). William J. Gedney’s Lue dialects : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Hudak, Thomas J. (Ed.). (1997). William J. Gedney’s Tai dialects : glossaries, texts, and translations. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Hudak, Thomas J. (Ed.). (2001). Minot's White Tai: a phonetic dictionary. Tempe, AZ: Program for South and Southeast Asian Studies, Arizona State University. Hudak, Thomas J. (Ed.). (2008). William J. Gedney’s Comparative Tai source book. Honolulu: University of Hawaii Press. Huffman, Franklin E. (1972). The boundary betwen monosyllable and disyllable in Cambodian. Lingua, 29, 54-66. Inkelas, Sharon. (1989). Prosodic constituency in the lexicon. Unpublished Ph.D. dissertation, Stanford University, Palo Alto. Institute of Language and Culture for Rural Development, Mahidol University, & Kam-Tai Institute, Central University for Nationalities. (2001). BouyeiChinese-English-Thai dictionary. Nakhon Pathom, Thailand: Institute of Language and Culture for Rural Development, Mahidol University. Institute of Language and Culture for Rural Development, Mahidol University, & Thailand and the Kam-Tai Institute, Central University for Nationalities, People’s Republic of China. (2003). Sui-Chinese-Thai-English Dictionary. Nakhorn Pathom, Thailand: Institute of Language and Culture for Rural Development, Mahidol University. 402 International Phonetic Association (Ed.). (1999). Handbook of the International Phonetic Association: A guide to the use of the International Phonetic Alphabet. Cambridge; New York: Cambridge University Press. Iverson, Gregory K., & Salmons, Joseph C. (1992). The phonology of Proto-IndoEuropean *DVD root structure constraint. Lingua, 87, 290-320. Jonsson, Nanna L. (1991). Proto-Southwestern Tai. Unpublished Ph.D. dissertation, State University of New York at Albany, Albany, NY. Jrang, Ja Kuang, Ja Wi, Ja Dai, Ja Ngai, & Fuller, Eugene. (1983). Chru vocabulary. Manila: Summer Institute of Linguistics. Karlgren, Bernhard. (1964). Grammata serica recensa. Stockholm: Museum of Far Eastern Antiquities. Kasisopa, Benjawan. (2003). Phonological study and genetic classification of Daraang (Palaung) spoken at Nawlae Village, Fang District, Chiangmai Province. Unpublished M.A. thesis, Chulalongkorn University, Bangkok. Kingston, John. (2004). Mechanisms of tone reversal. In S. Kaji (Ed.), Proceedings of the Third Symposium on Cross-linguistic Studies of Tone (pp. 57-120). Tokyo: Institute of the Study of the Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies. Kingston, John, & Diehl, Randy L. (1994). Phonetic knowledge. Language, 70, 419454. Kingston, John, & Solnit, David B. (1989). The inadequacy of underspecification. In J. Carter & R.-M. Déchaine (Eds.), Proceedings of the 19th Meeting of the Northeast Linguistic Society, 264-278. (pp. 264-278). Amherst, MA: University of Massachusetts, Amherst. Kong, Qingming. (1987). Influence of tones upon vowel duration in Cantonese. Language and Speech, 30, 387-399. Kosaka, Ryuichi. (1992). Tentative de reconstruction d'un proto-saek-chuang et comparaison de son vocabulaire avec l'ancien siamois. Unpublished M.A. thesis, Tokyo University of Foreign Studies, Tokyo. Kosaka, Ryuichi. (1997). A phonology of the Bu Dai (Thu Lao) language. Journal of Asian and African Studies, 53, 137-153. 403 Kosaka, Ryuichi. (2002). On the affiliation of Miao-Yao and Kadai –possibility of forming Miao-Dai linguistic family. Mon-Khmer Studies, 32, 71-100. Kosaka, Ryuichi. (2007). Consonantal lenition and vocalic transfer in Kadai languages: with disyllabic proto-Be-Tai as supporting evidence. In I. Shoichi, A. Simpson, K. L. Adams & P. Sidwell (Eds.), Papers from the 13th Annual Meeting of the Southeast Asian Linguistics Society 2003 (pp. 93-103). Canberra: Pacific Linguistics. Kreitman, Rina. (2008). Phonetics and phonology of onset clusters: the case of Modern Hebrew. Unpublished Ph.D. dissertation, Cornell University, Ithaca, NY. Kullavanijaya, Pranee. (1992a). Photchananukrom Chuang Tai-Thai [Southern Zhuang-Thai dictionary]. Bangkok: Faculty of Arts, Chulalongkorn University. Kullavanijaya, Pranee. (1992b). A study of some two-syllabled words in Thai. In PanAsiatic linguistics: Proceedings of the Third International Symposium on Language and Linguistics (pp. 651-656). Chulalongkorn University, Bangkok, Thailand. Kullavanijaya, Pranee. (2001). Phasa Thai Nuea [The Tai Nüa language]. Bangkok: Faculty of Arts, Chulalongkorn University. Kullavanijaya, Pranee, & L-Thongkum, Theraphan. (2000). Linguistic criteria for determining Tai ethnic groups: case studies on Central and South-Western Tais. In S. Burusphat (Ed.), Proceedings : the International Conference on Tai Studies, July 29-31, 1998. (pp. 273-298). Bangkok: Institute of Language and Culture for Rural Development, Mahidol University. L'Écôle Française d'Extême-Orient. (1938). linguistique.Unpublished manuscript, Hanoi. EFEO Questionaire L-Thongkum, Theraphan. (1989). An acoustic study of the register complex in Kui (Suai). Mon-Khmer Studies, 15, 1-19. L-Thongkum, Theraphan. (1991). An instrumental study of Chong registers. In J. H. C. S. Davidson (Ed.), Austroasiatic languages: Essays in honour of H. L. Shorto (pp. 141-160). London: School of Oriental and African Studies. 404 L-Thongkum, Theraphan. (1992). A preliminary reconstruction of Proto-Lakkja (Cha Shan Yao). Mon-Khmer Studies, 20, 57-90. L-Thongkum, Theraphan. (1997). Implications of the relation of proto-voiced plosives and fricatives in the Dai Tho language of Yunnan Province for a theory of tonal development and Tai language classification. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: The Tai branch (pp. 191-220). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. L-Thongkum, Theraphan. (2001). Languages of the tribes in Xekong Provice Southern Laos : a foundation for research and development. Bangkok: Thailand Research Fund. L-Thongkum, Theraphan. (2002). Old Tai Dam (Black Tai) and the meaning of ambiguous words in Modern Thai elaborate expressions. Paper presented at the Conference Name|. Retrieved Access Date|. from URL|. L-Thongkum, Theraphan (Ed.). (1978). Siang Lae Rabop Siang Nai Phasa Thai Thin Tai Changwat Sura Sot Thani 19 Amphoe [Sounds and sound systems of Southern Thai spoken in the 16 amphoes of Suratthani province]. Bangkok: Indegeneous Languages of Thailand Project, Chulalongkorn University. Ladefoged, Peter, & Maddieson, Ian. (1996). The sounds of the world’s languages. Oxford, UK ; Cambridge, MA: Blackwell Publishing. Lee, Ernest. (1966). Proto-Chamic phonologic word and vocabulary. Unpublished Ph.D. dissertation, Indiana University, Bloomington. Lehiste, Ilse. (1970). Suprasegmentals. Cambridge, MA: MIT Press. Levin, Juliette. (1985). A Metrical Theory of Syllabicity. Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Li, Fang-kuei. (1940). Lóngzhōu tǔyǔ [The Tai dialect of Lungchow]. Shāngwù yìnshū guǎn. Li, Fang-kuei. (1960). A tentative classification of Tai dialects. In S. Diamond (Ed.), Culture in history: Essays in honor of Paul Radin (pp. 951-959). New York: Columbia University Press. 405 Li, Fang-kuei. (1964). The phonemic system of the Tai Lü language. Bulletin of the Institute of History and Philology, Academia Sinica, 35, 7-14. Li, Fang-kuei. (1965). The Tai and the Kam-Sui languages. In G. B. Milner & E. i. J. A. Henderson (Eds.), Indo-Pacific Linguistic Studies (Vol. Vol. I, pp. 148-179 ). Amsterdam: North Holland Publishing Co. Li, Fang-kuei. (1971). On Siamese jaai. Bulletin of the Institute of History and Philology, Academia Sinica, 42(3), 337-340. Li, Fang-kuei. (1973). Some dental clusters in Tai. Bulletin of the School of Oriental and African Studies, University of London, 36(2), 334-339. Li, Fang-kuei. (1976). Sino-Tai. Computational Analysis of Asian and African Languages, 3, 39-48. Li, Fang-kuei. (1977). Handbook of Comparative Tai. Honolulu: University of Hawai'i Press. Li, Fang-kuei. (1980). Shànggǔ yīn yánjiū [Researches on Old Chinese phonology]. Běijīng: Shāngwù yìnshū guǎn. Li, Fang-kuei. (1989). Proto-Tai *kh and *x-. In J. H. C. S. Davidson (Ed.), SouthEast Asian Linguisitics: Essays in honour of Eugénie J.A. Henderson (pp. 143146). London: School of Oriental and African Studies, University of London. Li, Jinfang. (1999). Bùyāngyǔ yánjiū [Studies in the Buyang Language]. Běijīng: Central University for Nationalities Press. Liang, Min, & Zhang, Junru. (1993). Dòng-Tái yǔzú sòngqì qīngsè yīn shēngmǔ de chǎnshēng hé fāzhǎn [The rise and development of voicless aspirated initials in Kam-Tai languages]. Mínzú Yǔwén(5), 10-15. Liang, Min, & Zhang, Junru. (1996). Dòng-Tái yǔzú gàilùn [Outline of Tai-Kadai languages]. Běijīng: China Social Sciences Press. Liberman, Mark. (1975). The intonational system of English. New York: Garland Publishing. Liu, Baoyuan. (1999). Hàn-Yáo Cídiǎn (Lājiā yǔ) [Chinese-Yao dictionary (Lakkja language)]. Chéngdū: Sichuan Nationalities Press. 406 Luo, Yongxian. (1996). Tonal irregularities in Tai revisited. Mon-Khmer Studies, 25, 69-102. Luo, Yongxian. (1997). The subgroup structure of the Tai languages: A historicalcomparative study. Berkeley: Project on Linguistic Analysis, University of California, Berkeley. Luo, Yongxian. (1999). A dictionary of Dehong, Southwest China. Canberra: Pacific Linguistics. Maddieson, Ian. (1984a). The effects of a voicing distinction in sonorants and their implication for a theory of tonogenesis. Journal of Phonetics, 12, 9-15. Maddieson, Ian. (1984b). Patterns of sounds. Cambridge; New York: Cambridge University Press. Manomaivibool, Prapin. (1975). Study of Sino-Thai lexical correspondences. Unpublished Ph.D. dissertation, University of Washington, Seattle. Manomaivibool, Prapin. (1976). Layers of Chinese loanwords in Thai. In T. W. Gething, J. G. Harris & P. Kullavanijaya (Eds.), Tai linguistics in honor of Fang-Kuei Li (pp. 179-184). Bangkok: Central Institute of English Language, Office of State Universities. Maran, La Raw. (1978). Burmese and Jingpho; a study of tonal linguistic processes. Urbana-Champagne: Center for Asian studies, University of Illinois. Maspéro, Henri. (1912). Études sur la phonétique historique de la langue annamite. Les initiales. Bulletin de l'Ecôle Française d'Extrême-Orient, 12, 1-126. Matisoff, James A. (1973). Tonogenesis in Southeast Asia. In L. M. Hyman (Ed.), Consonant types & tones (pp. 71-95). Los Angeles: The Linguistic Program, University of Southern California. Matisoff, James A. (1992). Siamese jaai revisted, or ¡Ay, madre!: a case study in multiple etymological possibilities. In C. J. Compton & J. F. Hartmann (Eds.), Papers on Tai languages, linguistics and literatures (pp. 111-117). DeKalb: Northern Illinois Center for Southeast Asian Studies Matisoff, James A. (2001). Genetic versus contact relationship: Prosodic diffusibility in South-East Asian languages. In A. Y. Aikhenvald & R. M. W. Dixon (Eds.), 407 Areal diffusion and genetic inheritance: problems in comparative linguistics (pp. 291-327). Oxford; New York: Oxford University Press. Matisoff, James A. (2006). English-Lahu lexicon. Berkeley: University of California Press. Mayrhofer, Manfred. (1986). Lautlehre (Segmentale Phonologie des Indogermanischen) (Vol. I). Heidelberg: Carl Winter. Mazaudon, Martine. (1977). Tibeto-Burman tonogenetics. Linguistics of the TibetoBurman area, 3(2), 1-123. McCarthy, John, & Prince, Alan. (1986). Prosodic morphology. manuscript. University of Massachusetts, Amherst, and Brandeis University. Michaud, Alexis. (2004). Final consonants and glottalization: new perspectives from Hanoi Vietnamese Phonetica, 61(2-3)(2-3), 119-146. Morén, Bruce, & Zsiga, Elizabeth. (2006). The lexical and post-lexical phonology of Thai tones. Natural Language and Linguistic Theory, 24(1), 113—178. Morey, Stephen. (2004). Tonal change in the Tai languages of Northeast India. Linguistics of the Tibeto-Burman Area, 28(2). Morey, Stephen. (2005). The Tai languages of Assam: A grammar and texts. Canberra: Pacific Linguistics. Nacaskul, Kanchana. (1978). The syllabic and morphological structure of Cambodian words. Mon-Khmer Studies, 7, 183-200. Nespor, Marina, & Vogel, Irene. (1986). Prosodic Phonology. Dordrecht, Holland; Riverton, USA: Foris Publications. Nguyễn, Văn Lợi, & Edmondson, Jerold A. . (1998). Tones and voice quality in modern northern Vietnamese: Instrumental case studies. Mon-Khmer Studies, 28, 1-18. Nishida, Tatsuo. (1954). Tonēmu ni yoru Tai shogo hikaku gengogakuteki kenkyū [Comparative linguistic studies of the Tai languages based on tonemes]. Gengo Kenkyū, 25, 19-46. 408 Nishida, Tatsuo. (1955). Makku-Sui go to kyōtsu go hikaku gengogakuteki kenkyū [Mak-Sui and Common Tai]. Gengo Kenkyū, 28, 30-62. Nishida, Tatsuo. (1975a). Common Tai and Archaic Chinese. Studia Phonologica (2-12). Nishida, Tatsuo. (1975b). On the developments of tones in Tibetan. Acta Asiatica, 25, 43-55. Norquest, Peter. (2007). A phonological reconstruction of Proto-Hlai. Unpublished Ph.D. dissertation, University of Barcelona, Tucson, AZ. Ohala, John J. (2003). Phonetics and historical phonology. In B. D. Joseph & R. D. Janda (Eds.), The handbook of Historical Linguistics (pp. 669-686). Malden, MA: Blackwell Publishing. Ostapirat, Weera. (1995). Notes on Laha final -l. Linguistics of the Tibeto-Burman Area, 18(1), 173-181. Ostapirat, Weera. (2000). Proto-Kra. in Linguistics of the Tibeto-Burman Area, 23(1), 1-251. Ostapirat, Weera. (2004). Proto-Hlai sound system and lexicons. In Y.-c. Lin, F.-m. Hsu, C.-c. Lee, J. T.-S. Sun, H.-f. Yang & D.-a. Ho (Eds.), Studies on SinoTibetan languages: Papers in honor of Professor Hwang-cherng Gong on his seventieth birthday (pp. 121-175). Taipei: Institute of Linguistics. Academia Sinica. Ostapirat, Weera. (2005). Kra-Dai and Austronesian: Notes on phonological correspondences and vocabulary distribution. In L. Sagart, R. Blench & A. Sanchez-Mazas (Eds.), The peopling of East Asia: putting together archaeology, linguistics and genetics (pp. 107-131). London; New York: RoutledgeCurzon. Ostapirat, Weera. (2007). Proto-Tai velars and postvelars. In J. G. Harris, S. Burusphat & J. E. Harris (Eds.), Studies in Tai and Southeast Asian Linguistics (pp. 293300). Bangkok: Ekphimthai Ltd. Parker, Stephen. (2002). Quantifying the Sonority Hierarchy. Unpublished Ph.D. dissertation, University of Massachusetts Amherst., Amherst. Petsuk, Rasi. (1978). General Chracteristics of the Khün language. Unpublished M.A. thesis, Mahidol University, Nakhorn Pathom. 409 Pham, Andrea Hoa. (2003). Vietnamese tone : a new analysis. New York: Routledge. Phothiban, Phuthachart, & Trongdi, Thananan. (1998). Sathana Khong Phasa Tak Bai Nai Phasa Thai Thin [The Position of Tak Bai Tai in Tai Dialects (Phase 2)] (Research report). Pattani: Faculty of Humanties, Prince of Songkhla Univeristy. Pittayaporn, Pittayawat. (2005a). Moken as a Mainland Southeast Asian language. In A. Grant & P. Sidwell (Eds.), Chamic and Beyond (pp. 189-210). Canberra: Pacific Linguistics. Pittayaporn, Pittayawat. (2005b). Phu Thai Kapong. fieldnote. Cornell University. Pittayaporn, Pittayawat. (2005c). Tone, minor syllable, and infixation in Kammu. Paper presented at the 11th annual Midcontinental Workshop on Phonology, University of Michigan, Ann Arbor, Michigan. November 4-6, 2005. Pittayaporn, Pittayawat. (2006). Giay of Sapa. fieldnote. Cornell University. Pittayaporn, Pittayawat. (2007a). A chronology-sensitive approach to subgrouping: The case of Southwestern Tai. Paper presented at the Conference Name|. Retrieved Access Date|. from URL|. Pittayaporn, Pittayawat. (2007b, August 6-10, 2007). Directionality of tone change. Paper presented at the 16th International Congress of Phonetic Sciences, Saarbrücken, Germany. Pittayaporn, Pittayawat. (2007c). Tay of Bao Yen. fieldnote. Cornell University. Pittayaporn, Pittayawat. (2007d). Tay of Cao Bang. fieldnote. Cornell University. Pittayaporn, Pittayawat. (2007e). Tay of Sapa. fieldnote. Cornell University. Pittayaporn, Pittayawat. (2008a). Lakkja. fieldnote. Cornell University. Pittayaporn, Pittayawat. (2008b). Proto-Southwestern Tai: A New Reconstruction. Paper presented at 18th Annual Meeting of the Southeast Asian Linguistics Society, Universiti Kebangsaan Malaysia, Bangi. May 21-22, 2008. Pittayaporn, Pittayawat. (2008c). Zhuang dialect of Shangsi. fieldnote. Cornell University. 410 Pittayaporn, Pittayawat. (to appear-a). Establishing relative chronology of Palaung sound changes using Tai loanwords. Mon-Khmer Studies, 39. Pittayaporn, Pittayawat. (to appear-b). Proto-Southwestern-Tai revised: A new reconstruction. Journal of the Southeast Asian Linguistics Society. Pulleyblank, Edwin G. (1962). The consonantal system of Old Chinese. Asia Major, 9, 58-144, 206-165. Pulleyblank, Edwin G. (1984). Middle Chinese : a study in historical phonology. Vancouver: University of British Columbia Press. Pulleyblank, Edwin G. (1991). Lexicon of reconstructed pronunciation in Early Middle Chinese, Late Middle Chinese, and Early Mandarin. Vancouver: University of British Columbia Press. Qin, Xiaohang. (1992). Zhuàngyǔ liàngcí pou4 de láiyuán [Origin of Zhuang classifier pou4]. Mínzú Yǔwén(4). Ratliff, Martha S. (1992). Meaningful tone : a study of tonal morphology in compounds, form classes and expressive phrases in White Hmong. DeKalb, IL: Northern Illinois University, Center for Southeast Asian Studies. Rhee, Seok-Chae. (2003). Onset-coda asymmetries in Mishmi and other Southeast Asian languages. In K. L. Adams, T. J. Hudak & F. K. Lehman (Eds.), Papers from the Seventh Annual Meeting of the Southeast Asian Linguistics Society (pp. 137-152). Tempe, AZ: Program for Southeast Asian Studies, Arizona State University. Ross, Peter. (1996). Dao Ngan Tay: a B-language in Vietnam. Mon-Khmer Studies, 25, 133-139. Rungrueangsri, Udom. (1991). Photchananukrom Lanna - Thai Chabap Mae Fa Luang [Mae Fah Luang's Lanna-Thai Dictionary]. Bangkok: มูลนิธิแม่ฟ้ าหลวง [Mae Fah Luang Foundation. Sagart, Laurent. (1986). On the departing tone. Journal of Chinese Linguistics, 14(1), 90-113. Sagart, Laurent. (1988). Glottalized tones in China and South-East Asia. In D. Bradley, E. J. A. Henderson & M. Mazaudon (Eds.), Prosodic analysis and 411 Asian linguistics: to honour R. K. Sprigg (pp. 83-93). Canberra: Pacific Linguistics. Sagart, Laurent. (1993). Austronesian finals and the origin of Chinese tones. In J. A. Edmondson & K. Gregerson (Eds.), Tonality in Austronesian languages (pp. 41-60). Honolulu: University of Hawai'i Press. Sagart, Laurent. (1999). The roots of Old Chinese. Amsterdam; Philadelphia: John Benjamins Pub. Co. Sagart, Laurent. (2004). The higher phylogeny of Austronesian and the position of Tai-Kadai. Oceanic Linguistics, 43(2), 411-444. Sagart, Laurent. (2005). Tai-Kadai as a subgroup of Austronesian. In L. Sagart, R. Blench & A. Sanchez-Mazas (Eds.), The peopling of East Asia: putting together archaeology, linguistics and genetics (pp. 177-181). London; New York: RoutledgeCurzon. Sarawit, Mary E. (1973). The Proto-Tai vowel system. Unpublished Ph.D. dissertation, University of Michigan, Ann Arbor. Sarawit, Mary E. (1979). A sketch of a dialect of Mae Sot. In T. Gething & D. L. Nguyen (Eds.), Papers in Southeast Asian Linguistics No.6, Tai Studies in honor of William J. Gedney (pp. 75-84). Canberra: Pacific Linguistics. Sastry, G. Devi Prasada. (1984). Mishmi grammar. Mysore: Central Institute of Indian Languages. Saul, Janice E., & Freiberger Wilson, Nancy. (1980). Nùng grammar. Dallas, TX: Summer Institute of Linguistics ; Arlington : University of Texas at Arlington. Schlegel, Gustav. (1902). Siamese studies. T'oung Pao, ser. 2: vol. 2(supplement), 1128. Schmidt, Wilhelm. (1926). Die Spracfamilien und Sprachenkreise der Erde. Heidelberg. Schuessler, Axel. (2007). ABC etymological dictionary of Old Chinese. Honolulu: University of Hawai’i Press. Schuessler, Axel. (2009). Mininal Old Chinese and Later Han Chinese: A Companion to Grammatica Serica Recensa. Honolulu: University of Hawai'i Press. 412 Selkirk, Elizabeth. (1981). On the nature of phonological representation. In J. Anderson, J. Laver & T. Myers (Eds.), The cognitive representation of speech (pp. 379-388). Amsterdam: North Holland. Selkirk, Elizabeth O. (1984). On the major class features and syllable theory. In M. Aronoff & R. T. Oehrle (Eds.), Language sound structure (Vol. 107-136). Cambridge, MA: MIT Press. Shaw, Patricia. (1994). The prosodic constituency of minor syllable. In E. Duncan, D. Farkas & P. Spaelti (Eds.), Proceedings of West Coast Conference on Formal Linguistics 12 (pp. 117-132). Stanford: Stanford Linguistics Association. Shorto, Harry. (1960). Word and syllable pattern in Palaung. Bulletin of the School of Oriental and African Studies, University of London, 23(3), 544-557. Shorto, Harry. (2006). A Mon-Khmer comparative dictionary. In P. Sidwell, D. Cooper & C. Bauer (Eds.). Canberra: Pacific Linguistics. Silverman, Daniel, Blackenship, Barbara, Kirk, Paul, & Ladefoged, Peter. (1995). Phonetic structures in Jalapa Mazatec. Anthropological Linguistics, 37, 70-88. Simmonds, E. H. S. (1965). Notes on some Tai dialects of Laos and neigbouring regions. Lingua, 14, 133-147. Sloan, Kelly. (1988). Bare-consonant reduplication: Implications for a prosodic theory of reduplication. In H. Borer (Ed.), Proceedings of West Coast Conference on Formal Linguistics 7 (Vol. 319-330). Stanford: Stanford Linguistics Association. Steriade, Donca. (1982). Greek prosodies and the nature of syllabification. Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Stevens, Kenneth N., Keyser, S. Jay, & Kawasaki, Haruko (1986). Toward a phonetic and phonological theory of redundant features. In J. S. Perkell & D. H. Klatt (Eds.), Invariance and variability in speech processes (pp. 426-449). Hillsdale, NJ: Lawrence Erlbaum Associates. Strecker, David. (1979). A preliminary typology of tone shapes and tonal sound changes in Tai: the Lān Nā A-tones. In T. L-Thongkum, V. Panupong, P. Kullavanijaya & K. M. R. Tingsabadh (Eds.), Studies in Tai and Mon-Khmer phonetics and phonology in honour of Eugénie J.A. Henderson. Bangkok: Chulalongkorn University Press. 413 Strecker, David. (1983). Proto-Tai vowels revisited: a comparison and critique of the work of Sarawit and Li. Linguistics of the Tibeto-Burman area, 7(2). Strecker, David. (1984). Proto-Tai personal pronouns. Unpublished Ph.D. dissertation, University of Michigan, Ann Arbor, MI. Strecker, David. (1985). The classification of the Caolan languages. In S. Ratanakul, D. Thomas & S. Premsrirat (Eds.), Southeast Asian Linguistic Studies presented to André-G. Haudricourt (pp. 479-492). Nakhorn Pathom: Institute of Language and Culture for Rural Development, Mahidol University. Strecker, David. (1988). Gedney's Puzzle in Kam-Sui. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: Linguistic studies beyond Tai (pp. 107128). Arlington, TX: The Summer Institute of Linguistics and the University of Texas at Arlington. Svantesson, Jan-Olof. (1983). Kammu phonology and morphology. Lund: CWK Gleerup. Svantesson, Jan-Olof. (1991). Hu – a language with unorthodox tonogenesis. In J. H. C. S. Davidson (Ed.), Austroasiatic languages: Essays in honour of H. L. Shorto (pp. 67-79). London: School of Oriental and African Studies. t'Hart, Johan, Collier, René, & Cohen, Antoine. (1990). A perceptual study of intonation: An experiment-phonetic approach to speech melody. Cambridge: Cambridge University Press. Tài, Nguyyn Văn. (1982). Ngữ âm tiing Mường qua các phương ngôn [Muong phonology across dialects]. Hà NNi: Nguyyn Văn Tài. The Royal Institute. (2003). Photchananukrom Chabap Ratchabandit Sathan Pho So 2542 [The Royal Institute's Dictionary 2002 ]. Bangkok: Nanmee Book Publications. Theranon, Phaninthara. (2002a). Changes in phonetic characteristics of Thai Tone 4. Warasarn Aksorasat, Silpakorn University, 24(1-2), 188-209. Theranon, Phaninthara. (2002b). Thai tones in the reigns of King Rama VI and King Rama IX. Journal of Language and Linguistics, 21(1), 32-46. Thisaphong, Arilak. (1985). Rabop Siang Phasa Ka Loeng Thi Ban Dong Mafai Changwat Sakon Nakhon [The Phonological description of Kaloeng at Ban 414 Dong Mafai, Sakonnakhorn Province]. Unpublished M.A. thesis, Mahidol University, Nakhorn Pathom. Thompson, Laurence C. (1965). A Vietnamese reference grammar Seattle: University of Washington Press. Thurgood, Graham. (1988). Notes on the reconstruction of Proto-Kam-Sui. In J. A. Edmondson & D. B. Solnit (Eds.), Comparative Kadai: Linguistic studies beyond Tai (pp. 179-218). Dallas: Summer Institute of Linguistics and the University of Texas at Arlington Publications in Linguistics. Thurgood, Graham. (1996). Language contact and the directionality of internal drift: the development of tones and registers in Chamic. Language, 71(1), 1-31. Thurgood, Graham. (1999). From ancient Cham to modern dialects : two thousand years of language contact and change. Honolulu: University of Hawai'i Press. Thurgood, Graham. (2002a). A comment on Gedney's proposal for another series of voiced initials in proto-Tai. . In R. B. Bauer (Ed.), Studies in Southeast Asian Languages (pp. 169-183). Canberra: Pacific Linguistics. Thurgood, Graham. (2002b). Vietnamese and tonogenesis: revising the model and the analysis. Diachronica, 19(2). Thurgood, Graham. (2005). The tones from Proto-Chamic to Tsat [Hainan Cham]: Insights from Zheng 1997 and from summer 2004 fieldwork. In A. Grant & P. Sidwell (Eds.), Chamic and beyond: Studies in Mainland Austronesian languages (pp. 247-271). Canberra: Pacific Linguistics. Thurgood, Graham. (2007). Comments on Gedney's proposal for another series of initials in Proto-Tai revisited. In J. G. Harris, S. Burusphat & J. E. Harris (Eds.), Studies in Tai and Southeast Asian Linguistics (pp. 241-262). Bangkok: Ekphimthai Ltd. Trask, R. L. (2000). The dictionary of historical and comparative linguistics. Chicago: Fitzroy Dearborn. Trongrat, Phutphong. (1998). A Comparative Phonological Study of Thai Phuan at Suphanburi and Sukhothai Provinces. Unpublished M.A. Thesis, Mahidol, Nakhorn Pathom. 415 Umar, Umaiyah Haji. (2003). A classification of Thai-Kedah dialects using phonological characteristics. Bangkok: Nutcha Printing Limited Partnership. Ussishkin, Adam. (2005). A fixed prosodic hierarchy of nonconcatenative templatic morphology. Natural Language and Linguistic Theory, 23, 169-218. van der Hulst, Harry, & van de Weijer, Jeroen. (1995). Vowel harmony. In J. A. Goldsmith (Ed.), The handbook of Phonological Theory (pp. 495-534). Cambridge, MA: Blackwell. Varasarin, Uraisi. (1984). Les éléments khmers dans la formation de la langue siamoise. Paris: Société d’études linguistiques et anthropologiques de France. Venneman, Theo (1972). On the Theory of Syllabic Phonology. Linguistische Berichte, 18, 1-18. Wang, Stephen S. (1966). Phonology of Chinese loanwords in a northern Tai dialect. Unpublished Ph.D. dissertation, University of Washington, Seattle. Wayland, Ratree, & Guion, Susan G. (2005). Sound changes following the loss of /r/ in Khmer: a new tonogenetic mechanism. Mon-Khmer Studies, 35, 55-82. Wichiankhiaw, Arunrat et al. (1996). The Northern Thai Dictionary of Palm-Leafs Manuscripts (2nd ed.). Chiangmai, Thailand: Samnakphim Tratsawin. Wolff, John U. (in press). Proto-Austronesian Phonology and glossary. Wulff, K. (1934). Chinesisch und Tai : sprachvergleichende Untersuchungen. København [Copenhagen]: Levin und Munksgaard, Ejnar Munksgaard. Wulff, K. (1942). Uber das Verhältnis des Malayo-Polynesischen zum Indochinesischen. København [Copenhagen]: Levin und Munksgaard. Wurm, S. A., & Wilson, B. (1975). English finderlist of reconstructions in Austronesian languages (Post-Brandstetter). Canberra: Department of Linguistics, Research School of Pacific Studies, Australian National University. Xing, Gongwan. (2000). Hónghé Shàngyóu Dǎiyǎ yǔ [Tai Ya of the Upper Reach of the Red River]. Běijīng: Language and Literature Publishing. 416 Xu, Ching X, & Xu, Yi. (2003). Effects of consonant aspiration on mandarin Tones. Journal of the International Phonetic Association, 33, 165-181. Xu, Yi. (1997). Contextual tonal variations in Mandarin. Journal of Phonetics, 25, 61-83. Xu, Yi. (2001). Fundamental frequency peak delay in Mandarin. Phonetica, 58, 26-52. Xu, Yi. (2004). Understanding tone from the perspective of production and perception. Language and Linguistics, 5, 757-797. Yip, Moira. (2001). Tonal features, tonal inventories and phonetic targets. In C. Iten & A. Neeleman (Eds.), UCL Working Papers in Linguistics 13 (pp. 303-329). London: Department of Phonetics and Linguistics, Univerity College London. Yu, Alan. (to appear). Tonal effects on perceived vowel duration. In C. Fougeron & M. D'Imperio (Eds.), Laboratory Phonology 10. Berlin; New York: Mouton de Gruyter. Zec, Draga. (2005). Prosodic differences among function words. Phonology, 22, 77-112. Zec, Draga. (2007). The syllable. In P. de Lacy (Ed.), The Cambridge handbook of Phonology (pp. 161-194). Cambridge: Cambridge University Press. Zhang, Jie. (2002). The effects of duration and sonority on contour tone distribution. New York & London: Routledge. Zhang, Junru. (1995). Dòng-Tái yǔzú qīngchún yīn de chǎnshēng hé fāzhǎn [The rise and development of labiodental sounds in Kam-Tai languages]. Mínzú Yǔwén(1), 28-33. Zhang, Junru, Liang, Min, Ouyang, Jueya, Zheng, Yiqing, Li, Xulian, & Xie, Jianyou (1999). Zhuàngyǔ fāngyán yánjiū [Zhuang dialect research]. Chéngdū: Sichuan Nationalities Press. Zhou, Yaowen, & Luo, Meizhen. (2001). Dǎiyǔ fāngyán yánjiū [Dai dialect studies]. Běijīng: Minorities Press. 417