UNIVERSITY OF CALGARY

Explorations in the Phonological Reconstruction of Proto-Qiandong-Hmongic Onsets

by

Vincent Carveth

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF ARTS

DEPARTMENT OF LINGUISTICS

CALGARY, ALBERTA

JULY, 2012

©Vincent Carveth 2012 Library and Archives Bibliotheque et Canada Archives Canada

Published Heritage Direction du Branch Patrimoine de I'edition

395 Wellington Street 395, rue Wellington Ottawa ON K1A0N4 Ottawa ON K1A 0N4 Canada Canada Your file Votre reference ISBN: 978-0-494-91063-4

Our file Notre reference ISBN: 978-0-494-91063-4

NOTICE: AVIS: The author has granted a non­ L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lnternet, preter, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distrbute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non­ support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantiels de celle-ci substantial extracts from it may be ne doivent etre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis. Canada 11

UNIVERSITY OF CALGARY

FACULTY OF GRADUATE STUDIES

The undersigned certify that they have read, and recommend to the Faculty of Graduate Studies for acceptance, a thesis entitled "Explorations in the Phonological Reconstruction of Proto-Qiandong-Hmongic Onsets" submitted by Vincent Carveth in partial fulfilment of the requirements for the degree of Master of Arts in Linguistics.

Supervisor, Dr. Robert Murray, Linguistics

Dr. Stephen Winters, Department of Linguistics

Dr. Shu-ning Sci-ban, Department of Germanic, Slavic and East Asian Studies

Date Ill

ABSTRACT

This thesis concerns itself with the reconstruction of Proto-Qiandong-Hmongic onsets. It compares fieldwork data from two sources and 24 dialects of the Qiandong subgroup, spoken in China's Guizhou province, in order to synthesize a set of onsets for their common ancestor. The reconstruction has twofold relevance, in terms of both elucidating Hmong-Mien language history and bettering understanding of the typology of diachronic development of aspirated spirants, the most distinctive feature of the

Qiandong dialects. 34 onsets are reconstructed at the Proto-Qiandong-Hmongic level, including aspirated spirants, five distinct liquids, and prenasalized stops. The typology of fricative development is expanded by determining sources for multiple different aspiration contrasts in spirants, namely a reductive chain shift for the aspirated alveolars and palatals, palatalized bilabials for the aspirated labiodentals, and spirantizing aspirated liquids for the aspirated lateral fricative. Those developments are determined to have occurred prior to Proto-Qiandong-Hmongic in at least two prior stages. IV

ACKNOWLEDGEMENTS

Many people have been invaluable to this author in the creation of this thesis.

First, I owe a great debt of thanks to my advisor Dr. Robert Murray, whose patience and

counsel were critical throughout the research and writing process. The University of

Calgary Department of Linguistics as a whole similarly deserves my gratitude, for

providing me with education and funding allowing me to carry out such research. Alex

Mei and family were instrumental in kickstarting my research by translating Ma & Tai

(1956) for me as well as they could in the meantime while I searched for a professional

English translation; I would have been at a standstill for months without their efforts.

Finally, I want to thank my family, for providing me emotional and logistical support

throughout this two-year process. V

DEDICATION

I hereby dedicate this thesis to Maxwell Yang and Bee Vang, my Hmong language and culture teachers at the my undergraduate school, the University of Minnesota, for first acquainting me with the culture and the White Hmong language, piquing the interest that led me to this research. Ua tsaug ntau ntau, xibfwb. vi

TABLE OF CONTENTS

Abstract iii

Acknowledgements iv

Dedication v

Table of Contents vi

List of Tables ix

List of Figures xi

List of Abbreviations xii

1.0 Introduction 1

1.1 Data Used 3

1.2 Other Previous Research 6

1.3 Tone in Hmong-Mien 8

1.4 Typological Perspective On The Aspiration Contrast In Fricatives 11

1.5 Summary of Introductory Material 13

2.0 Language Orientation and Familial Context 13

2.1 Early Subgrouping Attempts 14

2.2 Subgrouping Anomalies 16

2.3 Conclusions Regarding Familial Context 21

3.0 Methodology and Theoretical Backdrop 22

3.1 The Comparative Method 23

3.2 Preference Laws 29 vii

3.3 Consultation of Existing Reconstructive Work 31

3.4 Lexical Diffusion 33

3.5 Exemplar Theory 35

3.6 Summary of Methodological and Theoretical Approach 39

4.0 Aspiration 39

4.1 Defining Aspiration 40

4.2 Aspiration and Voice 42

4.3 Summary of Aspiration Discussion 48

5.0 Plosives 49

5.1 Uvulars and Velars 51

5.2 Disappearance of the Aspiration Distinction in Tone D2 53

5.3 Palatal Stops and Affricates 56

5.4 Origin and Plausibility of Prenasalized Onsets 58

5.5 Summary of PQH Plosive Inventory and Changes 61

6.0 Nasals 61

6.1 Nasal Aspiration Contrast 63

6.2 Palatal-Velar Nasal Merger in Xuanwei 64

6.3 Nasal-Fricative Sequences 65

6.4 [l]-[p] Alternation 67

6.5 Summary of PQH Nasal Inventory and Changes 68

7.0 Non-Lateral Fricatives and Affricates 68 7.1 Qiandong Chain Shift 72

7.2 Labiodental Fricatives 82

7.3 Aspiration Mergers 85

7.4 Fricative-Aspirated Velar Stop Alternations 87

7.5 Voiced Velar Fricative Fronting 91

7.6 Summary of Fricative and Affricate Developments 93

8.0 Laterals 95

8.1 Reflexes Changed From PQH 97

8.2 Lateral Fricatives 100

8.3 Summary of Lateral Developments 107

9.0 Conclustion 107 ix

LIST OF TABLES

Table 1: Dialects Used 5

Table 2: Hmong-Mien Tone Categories 9

Table 3: Icelandic Sonorant Devoicing and Stop Deaspiration 46

Table 4a: Plosives in Ma & Tai 50, 78

Table 4b: Plosives in Purnell 51

Table 5: Uvular-Velar Stop Correspondences 52

Table 6: Aspiration Distinction Loss In Stops 54

Table 7a: Prenasalized Stop Series 57, 79

Table 7b: Non-prenasalized Stop Series 57, 79

Table 8a: Nasals in Ma & Tai 62

Table 8b: Nasals in Purnell 63

Table 9: Nasal Aspiration Merger 64

Table 10: Xuanwei Nasal Merger 65

Table 11: Unpredictable Nasal-Fricative Sequences 66

Table 12a: Fricatives and Affricates in Ma & Tai 69

Table 12b: Fricatives and Affricates in Purnell 71

Table 13: Shan Data 76

Table 14: Series 12 and 13, Reconstructed as Affricates 80

Table 15: Alveolar Affricates 81

Table 16: Tone D2 Fricative Aspiration Merger 86 X

Table 17: Tone A2 Fricative Aspiration Merger 86

Table 18: Aspiration Loss in PQH *f' 87

Table 19: Velar Fricatives and Aspirated Stops 87

Table 20: Aspirated Palatal Fricative Alternations 90

Table 21: Series 14 91

Table 22a: Laterals in Ma & Tai 95

Table 22b: Laterals in Purnell 96

Table 23: Lateral Deaspiration and Liquefaction Merger 97

Table 24: Jingxian Delateralization Merger 98

Table 25a: Alternating *lh Reflexes 99

Table 25b: Alternating *1 Reflexes 99

Table 26: Lateral Correspondences With Other Hmongic Branches 101 xi

LIST OF FIGURES

Figure 1: Purnell's Proto-Hmongic Family Tree 16

Figure 2: Mao & Meng's Proto-Hmongic Family Tree 18

Figure 3: Ratliffs Hmong-Mien Family Tree 20

Figure 4: Reconstructed PQH Plosives 61

Figure 5: Summary of PQH Nasal Inventory and Changes 68

Figure 6: Summary of PQH Onsets Reconstructed in Chapter 7 93

Figure 7: Lateral Reflex Attestation in Qiandong 96, 103, 106

Figure 8a: Pre-Palatalization State 106

Figure 8b: Liquid Palatalization 106

Figure 9: Summary of PQH Onsets Reconstructed in Chapter 8 107

Figure 10: PQH Onset Inventory 108 xii

LIST OF ABBREVIATIONS

EMC - Early Middle Chinese

LMC - Late Middle Chinese

MC - Middle Chinese

OC - Old Chinese

PH - Proto-Hmongic

PHM - Proto-Hmong-Mien

PQH - Proto-Qiandong-Hmongic 1

1.0 Introduction

This paper lays out a phonological reconstruction of Proto-Qiandong-Hmongic's consonant, vowel, and tone system. Proto-Qiandong-Hmongic (PQH) is the ancestor of the modern Qiandong dialects, which are spoken in southeastern Guizhou province in southwestern China by about 1.4 million people (Niederer 1998:51). The dialects are part of the Hmong-Mien language family, a language family currently thought to be unrelated to any other with about 6.3 million speakers, mostly in China (Ratliff 2010:3).

The speakers of the dialects in question are members of the Miao upland minority group in China. The group extends into the northern reaches of the Indochinese countries, in which they are often referred to with the autonym Hmong, in no small part because the speakers there consider the label Miao pejorative (Mortensen 2000). The particular dialects studied here are to an extent associated with the Black Miao/Hmong subgroup in Guizhou, although the correlation between language and ethnicity is quite loose (Niederer 1998:77). The Chinese ethnic label Miao, while associated with the

Hmongic family of languages, contains people who do not speak a Hmongic language at all, while some speakers of Hmongic languages are classified in China as Yao (the group that typically speaks the Mienic half of the greater language family)(Ratliff 2010:3). The problem of ethnicity in southern China is complex, and beyond the scope of this work, but a brief overview will hopefully help the reader situate the language.

The intent of this thesis is, through comparison of existing data on Qiandong 2 dialects, to reconstruct the onsets sufficiently to guide further work on the history of the

Hmong-Mien language family, of the Qiandong subgroup in particular, and of Southeast

Asia more broadly. More specifically, one of the distinctive traits of this subfamily is aspiration distinctions in fricatives in one or often more places of articulation. Such distinctions are extremely rare crosslinguistically and as such this work aims for broader typological relevance as well, through the reconstruction of the development of these contrasts. Furthermore, reconstructions at the bottom of the Hmong-Mien family, like this one, can inform and refine existing reconstructions at the higher levels of Proto-

Hmongic (PH) and Proto-Hmong-Mien (PHM), and thus enhance our understanding of

Hmong-Mien linguistic history.

The thesis is laid out as follows: Chapter 1 lays out all introductory content;

Chapter 2 discusses relevant matters of subgrouping and familial genetic context for

Qiandong, and their importance to its reconstruction; Chapter 3 contains information on the methodology and theoretical underpinnings of this thesis; Chapter 4 defines and discusses aspiration, with particular focus on aspiration of spirants, and on aspiration's interactions with voicing; Chapters 5-8 comprise the reconstruction itself, with each section detailing the development of a different sound class, with plosives, sonorants, fricatives, and lateral fricatives in succession. Finally, Chapter 9 summarizes and concludes. Each section will of course have subsections, and this chapter is no exception, with following sections on the source of the data used in the thesis, a summary of existing relevant research, and a typological note on the significance and relative cross-linguistic 3 rarity of aspirated fricatives.

Within Chapter 1, subsection 1.1 discusses the data used for reconstruction, subsection 1.2 lists previous historical Hmong-Mien work of note, subsection 1.3 is a note on tone categories in Hmong-Mien, subsection 1.4 provides a cross-linguistic, typological perspective on aspirated spirants, and subsection 1.5 summarizes previous sections.

1.1 Data Used

The language data used for this reconstruction comes almost exclusively from two sources. One is Purnell's (1970) Proto-Hmong-Mien reconstruction. Five of the dialects surveyed in that work are placed in his Eastern Miao group, a formulation equivalent to the Qiandong grouping in later work (see section 2.1 for details). Purnell (1970) lists phonological correspondences between the six dialects and their individual phonologies, as well as over a thousand purported PHM etyma, some of which have reflexes in one or more of the Qiandongic dialects surveyed.

The other source of data is Ma & Tai (1956), a Chinese fieldwork document on the phonologies of 17 further Qiandong dialects. While written in Chinese, it was translated in 1972 (Purnell 1972). It details the sound correspondences present between these dialects, and in the process lists several hundred vocabulary items.

As such, these documents provide a sufficient basis from which to begin 4

reconstruction, but are far from optimal in several ways. In particular, the vowel systems

are laid out in Ma and Tai (1956) in such a manner as to leave unclear which sound

correspondences are dominant and which are potential outliers. Purnell (1970) alleviates

this issue somewhat, but is sparse in how few of his etyma have Qiandongic reflexes, and

how many fewer yet have more than one such reflex. Furthermore, differences in

transcription are to be expected when borrowing from multiple sources. These

differences are typically systematic, and thus likely different interpretations of the same

sound; future fieldwork, refining the exact phonetic qualities of the data, would improve

the clarity of the data in this regard. Nevertheless, the documents drawn on here are sufficient to provide a basis for reconstructing PQH.

In Table 1 on page 5, all the dialects being used are listed, with their name in

Niederer (1998:301), which is the name that will be used in the rest of this thesis, the source they came from (Purnell (1970) or Ma & Tai (1956)), what they were called in that

work with its original spelling, and what subgroup of Qiand5ng the source places them

in. Niederer (1998) is used as the naming convention because it is the one work

containing all the dialects used in this reconstruction. While this thesis starts from a null

hypothesis regarding any groupings within Qiandong, as these proposed internal groupings are based on various traits, the source data often makes reference to these internal groupings and as such the connection between dialect and subgroup is valuable for reading clarity's sake. 5

Table 1: Dialects Used

Name Source Name in Source Subgroup Congjiang Ma & Tai Ts'ung chiang Jianhe Gaotongzhai Purnell MJC = Miao of Jung-chiang Jung-chiang Huangli Ma & Tai Huang li Leishan Jiaba Ma & Tai Chia pa Huangping Jinping Ma & Tai Chin p'ing Jianhe Jingxian Ma & Tai Ching hsien Jianhe Jinzhong Ma & Tai Chin chung Taijiang Jiuzhdu Ma & Tai Chiu chou Huangping Kaitang Purnell, Ma & Tai MLS = Miao of Lu- East A, Taijiang shan, K'ai tang Paiting Ma & Tai P'ai t'ing Jianhe Sandu Ma & Tai San tu Jianhe Sansui Ma & Tai San sui Jianhe ShTdongkou Purnell, Ma & Tai MTK = Miao ofTai- East A, Taijiang kung, Shih tung Taigongzhai Ma & Tai T'ai kung chai Taijiang Taiyong Ma & Tai T'ai yung Jianhe Wucha Ma & Tai Wu ch'a Taijiang Wuluo Ma & Tai Wu lo Leishan Xlnqiao Ma & Tai Hsin ch'iao Huangping Xuanwei Ma & Tai Ma chiang Lushan Yanghao Purnell MKL = Miao of East A K'ai-li Yong'an Ma & Tai Yung an Taijiang Zhenfeng Purnell MCF = Miao of East A Cheng-feng Zhenyuan Ma & Tai Chen yuan Taijiang ZhouxT Ma & Tai Chou ch'i Lushan 6

(Ma & Tai 1956:41, Niederer 1998:303-306, Purnell 1970:40)

1.2 Other Previous Research

Previous historical work in Hmong-Mien has exclusively looked at either higher levels or different branches. Much of it has taken place at the very top of the family tree, at the Proto-Hmong-Mien level. There have been three main reconstructions of PHM.

These are Purnell (1970), which is of course already in use here as a source of data, Wang and Mao (1995), and most recently Ratliff (2010), a reconstruction very much informed by the previous two. Also notable is Chang's (1976) reconstruction of Proto-Hmongic initials, which is affected significantly by Qiandongic evidence.

Chang also authored a set of papers (1953, 1966, 1972) about history of tones in

Hmong-Mien that "remains the definitive body of work on Hmong-Mien tonology today"(Ratliff 2010:4). Hmong-Mien's tones seem to have formed, like much of Asia's, in the manner described by Haudricourt (1954). Ratliff (2010:191) argues at length that the tonal languages in China and Southeast Asia developed their tones in parallel after lexically borrowing the phonetic basis for tone development en masse from a source language, quite possibly but not necessarily a Sinitic source. While tones play little role in the development of the onsets in Qiandong, this can to a great extent only be said definitively because these works have mapped out tonal development so well.

Three other sources notable for their utility in this paper are Niederer (1998), 7

Mortensen (2000), and Kwan (1966). Niederer (1998) is a reference work listing all published work, including Vietnamese and Chinese scholarship, done on Hmong-Mien historical phonology up until 1994. Moreover, it lists all of the known dialects, the work that has been done on each, and the alternate names they are known under. It even includes estimates of speaker numbers, as of the mid-1990s, for each major subgroup. It is in short a very detailed and helpful resource that has been consulted repeatedly in developing this thesis.

Mortensen (2000) is a painstaking examination of borrowings from Chinese languages into Hmongic, and most particularly into the well-studied Chuanqiandian dialects Hmong Daw/White Hmong and Mong Leng/Green Mong. While this is obviously not as helpful as it would be if it was a work on loans into Qiandong dialects, it still is quite helpful in several ways. For one, it gives an idea what words to flag as potential loans in Qiandong. Secondly, it provides a great deal of information about current understanding of Middle and Old Chinese from a Hmong-Mienicist's perspective, essential in identifying Chinese influences on Hmong languages.

Kwan (1966) is a synchronic account of the phonology of the ShTdongkou dialect of Qiandong. While it doesn't offer anything comparative in and of itself, it showcases a great deal of the dialect's vocabulary, and describes the vowel system in a systematic way, whereas the ShTdongkou vowel data in Ma & Tai (1956) is not presented in such a way as to make the phonemic system evident. 8

1.3 Tone in Hmong-Mien

Hmong-Mien languages, like most languages in China and Indochina, are tonal.

Tone does not seem to play a large role in consonantal development in Qiandong, but nevertheless tone has an effect on stops in section 5.2 and in fricatives in section 7.3.

Thus, a note on the tone categories, the relevant notation, and current understanding of

Hmong-Mien historical tonology is in order.

Tonal developments are perhaps the best understood part of historical Hmong-

Mien, with perhaps some of the best work on the topic being that of Chang (1953, 1966,

1972). This is largely because the tonal categories predominantly correspond to those in

Chinese, Tai, and Vietnamese, and as such are presumed to have originated in the same way. Indeed, Ratliff (2010:191) argues that the vocabulary from which tones developed was borrowed across Asia prior to tonogenesis, which would explain such a clean categorical correspondence. In any case, like its counterparts, tones in Hmong-Mien derive from syllable-final laryngeals (Haudricourt 1954). The loss of such consonants provided three different tones, shown below in (1).

(1) A(*-0) B(*-?) C(*-h)

A fourth category D arose later within Hmongic from the loss of final voiceless stops; many Mienic varieties retain these stops. Finally, each of these tones split based 9 on the voicing of initial consonants, producing the array of tones seen below. Both the numeric and alphanumeric monikers for each category are seen in the literature, so both are provided; the numeric labels have been used in the superscript for each word for brevity's sake in this work, but the alphanumeric category names are more succinctly informative of the category's origin, and will be used in discussion in this thesis.

Table 2: Hmong-Mien Tone Categories

A B C D Voiceless A1 (1) B1 (3) CI (5) D1 (7) Voiced A2 (2) B2 (4) : C2 (6) D2 (8) j (Ratliff 2010:184)

The consensus is currently that this split resulted in higher pitches for the syllables that had had voiceless onsets and lower pitches for the syllables that had had voiced onsets. However, things are more complicated than that. For instance, in a handful of languages this split based on voicing either did not occur at all or the obstruent distinction was not lost; YejTpo/Fuyuan in Chuanqiandian retains the original onsets and the original four tone categories, whereas Shimenkan, also Chuanqiandian, has split the tones without the loss of the voicing contrasts (Niederer 1998:109,116, Ratliff 2010:185). Additionally, not all modern Hmongic languages maintain higher pitches in the formerly voiceless tones than in the formerly voiced ones (Ratliff 2010:189). This suggests that some mechanism of tonal change enables their respective pitches to be interchanged. 10

While tone categories are remarkably stable over time, changes in category can and do continue after this. Later tonal splits in Hmongic are fundamentally local, both in terms of which languages are affected and in terms of not necessarily applying to all tone categories. Tone splits based on onset aspiration and on onset prenasalization, for instance, are attested, the former in the Chuanqiandian language Jiaotuo/Zongdi and the latter more generally in Xiangxi (Niederer 1998:246-248, Ratliff 2010:185). Moreover, most Hmongic languages have mergers involving two or more of the eight family-wide categories, one example being the merger of reflexes of categories D1 and B2 in White

Hmong (Ratliff 2010:185). However, all eight tones are attested, aside from such local phenomena, in Qiandong, indicating that all of the major Hmongic tone splits were completed prior to PQH.

Overall, though, tonal categories are quite stable things; aside from broad-based tone merger and split, the membership in a given category is more or less constant, and words that shared the same tone category as tones arose, words that prior to tonogenesis fell in the same class with regards to initial and final consonants, will continue to pattern together tonally. This uniformity is "why, of all aspects of Hmong-Mien historical phonology, tone was the first to be reconstructed"(Ratliff 2010:185). By comparison, the actual phonetic qualities associated with the categories are highly mutable, and for any given category their reflexes may differ widely across the languages of a family. For this reason, it is the category that is given with examples in this thesis and most other historical work on tone languages, rather than a phonetic label. 11

1.4 Typological Perspective On The Aspiration Contrast In Fricatives

As mentioned previously, the aspiration distinction is quite rare in fricatives.

Maddieson's (1984a) survey includes only three languages with such a contrast, and even

then it is only in [s] vs. [sh]. Some additional examples arise in the literature, and Jacques

(2011:2) undertakes an elaborate typological survey to uncover them. Most are Asian, and Sino-Tibetan in particular has several languages with the contrast, such as Burmese,

Sgaw Karen and several varieties of Tibetan. Jacques (2011:1) writes, notably, "The presence of aspirated fricatives in Asia is probably at least in part an areal development, as many of these languages [...] are spoken in contiguous zones." Late Middle Chinese has been reconstructed with aspirated fricatives, which may indicate a historical connection of some kind between the phenomenon's instantiations in various languages of Asia (Pulleyblank 1984a:63-9).

However, there are also instances of aspirated fricatives elsewhere, particularly in

Oto-Manguean (Kirk 1966, Knapp 1996, Pike & Pike 1947, Rensch 1976, Silverman et al. 1995), as well as isolated cases elsewhere, such as Ofo (Rankin 1988, 2004) and

Chumashan (Klar 1977:13-15). This indicates that an aspiration distinction has developed separately in different places, and that, while it can still be considered somewhat of a dispreferred characteristic (see chapter 2 for more on phonological preference), it should be considered a plausible one for the purposes of reconstruction.

Following in the tradition of Greenberg (1963), Jacques (2011:5), from gathered 12 typological evidence, suggests the implicational hierarchies shown in (2) below. If the phonemes on the left of the implicational mark are present, those on the right are expected to be as well. International Phonetic Alphabet notation is used here, as will be the case throughout this thesis.

h h 1 h (2) a. {9 , x , \\ P } => {G'\ §\ J*} => s b. f1 => sh

c. voiced aspirated fricatives => voiceless aspirated fricatives

Importantly for the purposes of this work, one of the languages Jacques (2011:5) draws upon is Yanghao, a Qiandong dialect. It appears in the data set used here as well due to its presence in Purnell (1970)'s dissertation. Yanghao has five different aspirated fricatives, the highest of any in Jacques (2011). However, as will be seen later, this is not aberrant within Qiand5ng.

Thus, reconstructing PQH will require explaining not only elaborate systems of aspirated fricatives but the most elaborate known. In this regard, there is distinct typological value in reconstructing PQH. While Jacques (2011) details several historical pathways that could create an aspiration contrast in a fricative, this is the first elucidation of the development of such a robust set of multiple aspirated fricatives. 13

1.5 Summary of Introductory Material

This thesis draws primarily on Ma & Tai (1956) and Purnell (1970) for data, while building on existing historical Hmong-Mien scholarship such as Niederer (1998) and

Ratliff (2010). The Qianddng languages, like the rest of Hmong-Mien, are tonal, with 8 tone categories present in the family, and the origins of those categories are remarkably well understood. The aspirated spirants in Qiandong are rarer cross-linguistically, but most common in Asia and far from unheard of worldwide.

2.0 Language Orientation and Familial Context

Any reconstructive effort on Qiandong needs first to delineate what exactly the

Qiandong group is, and as such a look at existing work on the structure of Hmongic is highly relevant here. Indeed, such an investigation suggests that Qiandong is a good place to begin reconstruction of the lower levels of the Hmong-Mien family. Subsection

2.1 details the earliest efforts at subgrouping, particularly those by Purnell, subsection 2.2 discusses anomalies in preexisting subgrouping efforts, and subsection 2.3 summarizes what this means for reconstruction attempts such as this thesis. 14

2.1 Early Subgrouping Attempts

The first attempt at elucidating the substructure of the Hmong-Mien family and the Hmongic branch in particular was by Ma Xueliang (1957), who recognized four different Hmongic subgroups, named for their geographic distribution as Eastern, Central,

Western, and Northern (Purnell 1970:38). This was followed by a Chinese Institute of

National Languages (1NL) (1970:38) study that kept Ma's Eastern and Central groups, naming them Xiangxl and Qiand5ng respectively after literary names for the geographical areas they were found in, while combining the Northern and Western groups into a single large Chuanqiandian branch (1970:38). Purnell (1970), however, distinguished four groups, and confusingly, gave them geographic names as Ma did, but different names to different groups. The Qiandong group, Ma's Central branch, is

Purnell's Eastern, and Ma's Eastern, Xiangxl, is Purnell's Northern(1970:38). He resubdivides Chuanqiandian into a Central and Western group, but these groupings are independent of Ma's groupings. Following the INL (1970) and Niederer (1998), the

Chinese names will be used here when possible to prevent confusion with regards to the differing geographical names.

According to Purnell, the main factors motivating Ma and the INL's groupings were shared vocabulary and a measure of mutual intelligibility. In contrast, Purnell focuses on the diachronic evidence. The first division he makes within Hmongic has to do with the lack of two classes of sounds, prenasalized obstruents on the one hand and 15 aspirated fricatives on the other. Purnell reconstructed a loss of prenasalized obstruents in Qiandong, and a loss of aspirated spirants elsewhere, and as such the Qiandong branch splits off higher in the tree than the others (1970:42). One of the Qiandong varieties in

Purnell's dissertation, which he terms Jung-chiang (Rongjiang), also lacks voiceless nasals and the aspiration distinction in voiceless fricatives.

This leaves the XiangxT dialects and Purnell's two Chuanqiandian groupings,

Central and Western. Purnell gives several reasons for not grouping these together. First is the resemblance of the finals of the XiangxT dialect in his study to those of Qiandong in general and those of Rongjiang particularly. Secondly, it also shares with Qiandong nasal reflexes for his reconstructed Proto-Hmongic voiced prenasalized obstruents (1970:41).

Purnell's use of only one dialect has limitations that he himself comments on: "Further data on dialects similar to Hua-Yuan is needed [...] since [XiangxT] is represented here by a single dialect," but the data at hand does suggest that XiangxT could not be part of a non-Qiandongic supergroup unless that supergroup had lost these features, preserved in

Qianddng, in all other descendants.

Purnell also uses evidence from tone categories to justify breaking up the other groups. Out of XiangxT, Central and Western, tone category A splits in XiangxT depending on initial voicing, category C splits in Western with a large subgroup of

Western, which he terms West A, splitting the other three categories as well, and finally

Central has no splits of this sort (1970:42).

The resulting structure has four major subgroups, two of which have 16 subgroupings of their own, West A in Western and East A in Qiandong. This is shown below in Figure 1. Niederer's names for the dialects are used, with Purnell's in parentheses, for clarity of comparison with their names elsewhere in this thesis.

Figure 1: Purnell's Proto-Hmongic Family Tree

Proto-Hmongic

Western XiangxT Central Qiandong /1\ I A A West A Shmienkan Gezheng LAyTping Shurwei JiSding East A Rongjiang (Wei-ning) (Kwang-shun) (Hua-yuan) (Lung-li) (Kwei-chu) (Jung-chiang)

(Purnell 1970:38)

2.2 Subgrouping Anomalies

The discovery of new information, however, on several previously understudied

Hmong-Mien speech varieties, rapidly complicated matters. Often these languages' place in the family tree was very much unclear. Mao & Meng (1982; cf. Strecker 1987), in deciding to place one such language, Ho-Ne, in its own branch rather than Mienic or 17

Hmongic, in the 1950s, arrange the family as seen on the following page in Figure 2.

While the position of Ho-Ne is still in question, with Chen (1984a) placing it in Mienic and Rat lift' (1998) placing it in Hmongic, the important innovation here is the reintegration of Purnell's Western and Central Hmongic into a single

Western/Chuanqiandian group, a configuration retained by subsequent scholarship

(Ratliff 2010, Niederer 2002).

The dialect group called Pu-Nu, another problematic group for subgrouping, is also evident in Figure 2 on the next page, where it is subdivided into a great many varieties by Mao and Meng (1982). There have been a variety of different opinions on the proper placing of Pu-nu. Mao & Meng (1982) place it as a sister to a Hmong group and daughter to Hmongic. Strecker (1987) and Niederer (2002:131) place it within

Western/Chuanqiandian Hmong, while LI (1995) seems to shy away from the same conclusion for purely ethnological reasons; Pu-nu's speakers are considered ethnic Yao in the Chinese system of ethnic classification, unlike the speakers of Hmong, who are considered Miao, and as such he places Pu-nu as a sister to a collective Hmong group, as shown below, while simultaneously remarking that, using Niederer's English paraphrase,

"this language behaves perfectly like a subdialect of Western Hmong" (Niederer

2004:137). 18

Figure 2: Mao & Meng's Proto-Hmong-Mien Family Tree

Hmong-Mjen

Hmongic Ho-neic Mjenic (midoyuzhi) (sheyuzhi) (ydoyuihl)

Hmong pu-nu ho-ne (midoyu) (bunuyu) (sheyu)

NA- E\W pu-nu pa-hng ng-nai iu-nuo kiong-nai mien/mun biau/tcau dzau-min unclassified tung-nu nu-nu pu-nuo nau-klau na-hraau

(Mao & Meng 1982, Niederer 2002:130)

Ratliff (1998:102) treats it as Chuanqiandian without elaborating on the matter.

As such, the current consensus view seems to be that it is within Chuanqiandian, contrary to Figure 2. However, the fact that there is still debate on this matter suggests that serious reconstructive work on Chuanqiandian is premature until its boundaries are better 19 defined.

A third major speech variety whose placement disrupts the diagram is Pa-Hng, a variety Mao & Meng originally categorized in Figure 2 as a Pu-nu dialect, which would, if the arguments for Pu-Nu as Chuanqiandian in the paragraph above hold, render Pa-Hng

Chuanqiandian as well. However, more fieldwork was done in the 90s, particularly by

Wang & Mao (1995), who posited it as a sister to a Hmong subgroup under Hmongic.

Niederer (2004:137) argues that Pa-Hng, while Hmongic, is "the most conservative

Hmongic language" , a view shared by Benedict (1986), who conjectured on the limited data available at the time (largely the notes of a French officer serving in North Vietnam in 1905) that it was either the first branch off of Hmongic or a third branch of Hmong-

Mien on its own. As such, Ratliffs (2010:3) tentative Hmong-Mien family tree, seen in

Figure 3 on the next page, which to my knowledge is the most recent attempt at subgrouping, holds that Pa-Hng branches off of Hmongic first, then Ho-Ne and its close relative Jiongnai separate from the mainstream Hmong languages just below that. Yet again, however, this issue speaks to the lack of clarity as to the bounds of the

Chuanqiandian group.

There is one yet further concern regarding Chuanqiandian. Johnson (2000:26) proposes a subdivision of Chuanqiandian he calls "Far-Western-Hmongic", roughly equivalent to Purnell's West-A subgrouping with more elaboration on the matter, but states that the details of the subgrouping will be dealt with in a later thesis. After twelve years, that thesis has not materialized, so this piece of the puzzle is tantalizingly 20 incomplete. Nevertheless, Chuanqiandian being largest of the three branches, with thirty dialects listed in Niederer (1998), these questions of internal substructure make a bottom- up reconstruction of its proto-language yet more difficult and arguably premature.

Figure 3: Ratliff s Hmong-Mien Family Tree

Hmong-Mien

Hmongic

Mien-Mun BiaoMin ZaoMin

Jiongnai/Ho-Ne

(Ratliff 2010:3)

Placing Pa-Hng and Ho-Ne higher in the tree, and Pu-Nu within Chuanqiandian, would seem to imply a valid subgroup for Hmong proper with three interior branches,

Chuanqiandian, XiangxT and Qiandong. Ratliffs (2010:3) tree, seen in Figure 3 above, does indicate this. Under Mien-Mun and the Core Hmongic subgroups, terminal nodes in the figure contain languages as divided by Ratliff. Thus, most if not all of the varieties in 21 this study would be considered dialects of Hmu language by Ratliff, but given the ambiguity between dialect and language, and the sparseness of existing Qiandong fieldwork, no attempt has been made in this work to subcategorize Qiandong at a level above the individual dialect.

However, she herself seems to have doubts. XiangxT seems to pattern similarly to

Pa-Hng in several key ways, and the data on XiangxT shows a great deal of conservatism

(Ratliff 2010:24). Whereas Hmongic typically preserves the first part of a Hmong-Mien diphthong if it preserves only part of the diphthong, both Pa-Hng and Qo Xiong, the variety Ratliff (2010) uses to represent XiangxT in her reconstruction, preserve the latter part instead. Also, Pa-Hng and Qo Xiong both preserve in two sets the effect of a lost coda consonant in a vowel quality distinction depending on whether the original syllable was open or closed. Thirdly, in another set the loss of the coda consonant from Hmong-

Mien creates a tone distinction. Finally, in a fourth set a Hmong-Mien rounded onglide causes vowel rounding which is preserved in Qo Xiong. This leads Ratliff to write that

"The archaism of Qo Xiong suggests that the structure of the Hmong-Mien family tree needs to be recalculated from the beginning, with no preconceptions about Hmongic language sub-grouping" (2010:24).

2.3 Conclusions Regarding Familial Context

In summary, out of the three core Hmongic subgroups, Chuanqiandian is a large 22 subfamily that has unclear bounds and internal divisions, and XiangxT's status as a subgroup of Core Hmongic is questionable due to commonalities with Pa-Hng. As such, neither is ready for reconstruction. However, Qiandong's boundaries are not contentious in the same way, nor is there any hints of internal subdivisions other than Purnell (1970)'s

East A, which was posited solely on the basis of the absence of aspirated fricatives in

Gaotongzhai (1970:41). This makes Qiandong comparatively ripe for phonological reconstruction. Moreover, a reconstruction of Proto-Qiandong onsets paves the way for future study of PQH rimes. The degree of PHM vowel information preserved in Proto-

Qiandong, compared to the degree seen in XiangxT and Pa-Hng, will have an impact on what relations should ultimately be assigned between the various subgroups. Finally, conclusions reached from reconstructing PQH can be applied to reassess PH and PHM.

3.0 Methodology And Theoretical Backdrop

It would be remiss not to discuss the method followed in reconstructing PQH in this thesis, as well as the theoretical assumptions underpinning the work; as with any intellectual endeavor, choices made at the stage of initial assumptions have a great deal of influence over the conclusions in historical linguistic reconstruction. This thesis follows the Comparative Method pioneered by the Neogrammarians, but assumes an exemplar theoretic framework for sound change and synchronic phonology. The implications of and reasons for each of these choices and the mechanics of their interaction are the 23 subject of the next few subsections. Subsection 3.1 describes the Comparative Method and its application, subsection 3.2 discusses Vennemann's Preference Laws and their gradient variant of markedness as used in this thesis, subsection 3.3 delineates when and how existing reconstructive work will be drawn upon, subsection 3.4 discusses lexical diffusion and the problems it presents for traditional Neogrammarianism, subsection 3.5 briefly explains how lexical diffusion and similar difficulties led exemplar theory to be chosen as a framework for this reconstruction, and finally subsection 3.6 recaps the previous subsections.

3.1 The Comparative Method

The approach to phonological reconstruction used in this thesis is the

Comparative Method, the standard approach to reconstructing protolanguage phonologies from daughter varieties made over the last two centuries. The details of this method's operation are laid out in this section.

The first relevant concept is the correspondence set. If, as the historical linguists of the 1800s known as the Neogrammarians postulated, sound change is fundamentally regular, then there should be solid, exceptionless correspondences between one sound in one daughter language and another in another daughter language, both coming from the same sound in the proto-language. Systematically mapping out the correspondences of sounds is critical to establishing a basis for phonological reconstruction (Hoenigswald 24

I960:119); it allows for later reconstruction of the phonemes of the ancestor language,

and the phonetic interpretation of those phonemes.

These correspondences, while exceptionless theoretically, in practice have

apparent exceptions. For instance, the daughter languages likely have borrowed some

words from other languages, or from each other if they are still in contact. These words

will not usually fit the expected pattern. Consider the words father and paternal in

English. They both come from the Proto-Indo-European root *pater, but differ in their

initial consonant. The reason becomes clear looking at Latin, where the corresponding

forms are pater and paternalis, respectively. Father has a fricative as a result of Grimm's

Law, but paternal is a borrowing of the Latin form, and Latin did not undergo the

processes subsumed under Grimm's Law, which only affected Proto-Germanic (Hock

1991:47).

Onomatopoeia can also disturb the regularity of correspondences. As an example,

the sound a dog makes in English is bow wow, while in German it is wau wau, yet there

is no common origin for the two. They are similar in sound merely because they are both

trying to mimic the same nonlinguistic sound. Moreover, cases of onomatopoeia tend to show resistance to regular sound changes, as the sound they are mimicking does not

change. Middle English pi:pen, the sound of bird chicks, should have changed to

Modern pipe [pajp] according to regular sound change rules, but chicks in English still

peep [pip] (1991:50).

There are other sorts of apparent exceptions, such as analogy, taboo, and so on. 25

However, these are all only technicalities; there are clear reasons for them to disobey the otherwise steadfast rule. More threatening to the Neogrammarian idea of regular sound change are posited cases of "lexical diffusion", a change gradually spreading through the vocabulary in a distinctly non-Neogrammarian manner since they assumed an across-the- board application of sound changes (Wang 1969). This will be addressed more fully in subsection 3.4.

In any case, the second step once sets of correspondences have been established is reconstructing the key phonological units of the ancestral language. These units have traditionally been phonemes or something like them, even if the structuralist frameworks of more recent phonology that define the term were in many cases not yet existent.

However, smaller articulatory units, such as gestures, or even individual muscle movements, may also be used in this capacity (Phillips 2002). This possibility is discussed further in section 3.5.

In the phonemic approach, which still constitutes the standard approach, changes can be categorized as splits and mergers. Mergers are fairly self-explanatory, as two phonemes of the ancestral language lose the distinctions between them. Splits come in two main types. Secondary splits, often referred to simply as splits, are those in which an allophonic difference in a single phoneme is phonologized, resulting in two daughter phonemes. In primary splits, a split in an existing phoneme results in a shift in some contexts to another existing phoneme. In other words, a split combined with merger changes certain sounds without changing the number of actual phonemes in the language 26 at that point (Hoenigswald 1960); this number can, of course, change later, as phonemes and phonological properties can be lost from a language without trace.

From Proto-Germanic to Old Germanic dialects other than Gothic and some runic inscriptions, a prime example of a merger takes place. A phonemic distinction between

*z and *r collapses, the *z disappearing and [r] becoming the universal form. The rule for this is shown in (3a) and examples in (3b) below.

(3a) *z > r (3b)

Proto-Germanic I *aiz Old English a:r !a:r Gloss 'metal' messenger (Hock 1991:54)

A good example of a primary split comes to us from the transition from Proto-

Indo-European to Proto-Germanic.(Hock 1991:53) In PIE, *d contrasted with *dh. In

Proto-Germanic, the contrast was instead between *t and *d/d. The change between the two happened according to the Grimm's Law chain shift, which followed the pattern listed below in (4), each sound moving one position to the right.

(4) bh > b > p > f dh > d > t > 0 gh > g > k > x

gwh > gw > > xw 27

However, in this case things are more complicated than that. In some contexts, which are still not entirely understood, the *dh became *d as expected, but in others it spirantized to *d (Ringe 2006:100). As there was already a *d generated by Grimm's

Law and Verner's Law from PIE */, no new phoneme was created by this split, some segments merely transitioned from one phoneme to another.

A representative example of a secondary split is thought to have occurred in Indo-

Iranian. Proto-Indo-European has at points been reconstructed with *k"' having a palatalized allophone *c before */ and *e and unpalatalized *k elsewhere. In Indo-

Iranian, PIE *e and *o merged into *a in virtually all environments. As a result, the *c and *k became contrastive before *a, and consequently they can be analyzed as separate phonemes (Hoenigswald 1960:94).

While the algebraic nature of this exercise necessitates that the actual phonetic qualities reconstructed be relatively vague, for anything resembling a real phonology to emerge from reconstruction some attempt must be made to hash out the phonetic properties of the reconstructed elements. Given a set of varying correspondences, there are several principles that can motivate the choice of one phonemic representation for the protolanguage over others. One is simply majority rule; if one phoneme is the dominant representation across the daughter languages, it makes a good choice for the protolanguage because relatively few changes have to be posited to explain the other 28 forms in the other daughter languages. This principle is reiterated below in (5).

(5) Majority Rule - A reflex common to more daughter varieties in the correspondence set is a better choice for reconstruction than one extant in fewer.

However, there are other principles in play. Changes might be suggested by the majority-rule measure that are phonetically unnatural, and these should be posited with caution proportional to their rarity. Also, since the mother language cannot logically have two of the same phoneme, if the aforementioned metrics suggest the same phoneme for more than one set of correspondences, at least one is going to have to be reconstructed as something else, most likely a phonetically similar segment. In other words, a reconstruction needs to be phonetically and typologically plausible, as the definitions in

(6) and (7) below are intended to indicate. Plausibility need not be defined at length for this thesis, but a more formalized notion of plausibility is covered in section 3.2.

(6) Phonetic Naturalness - A change that is phonetically plausible is a better choice for reconstruction than one that is not.

(7) Non-Duplication - A single phoneme and environment should not be constructed for more than one correspondence set. 29

3.2 Preference Laws

A conception of plausibility that works well for our purposes comes from

Preference Theory (Mailhammer, Restle & Vennemann in press). Their work is ultimately a development of Jakobson's markedness theory, but unlike most such theories of markedness, rather than a discrete contrast of marked and unmarked, Mailhammer et al's Preference Laws work on a graded scale of more to less preferred. That is to say, preference laws have the general form "X is more preferred than Y in terms of Z; Z is a gradable property of X and Y" (Mailhammer, Restle & Vennemann in press). This allows a certain fluidity while still holding onto the assumption that a change in the language's system is a local improvement in some parameter. Specifically, this relativization of markedness allows for entire scales in which structure B is marked compared to structure

C, but unmarked compared to structure A, which a simple notion of "marked" vs.

"unmarked" does not do.

An important starting notion in their work is Vennemann (2000:3)'s Synchronic

Maxim and Diachronic Maxim, shown below in (8) and (9) respectively. The Synchronic

Maxim, phrased another way, states that a language should not have less preferred structures unless it also has the related preferred structures. The hedge of "in general" is necessary, because at times gaps in the scale of preference can occur, as a result of phonetic impossibilities, language contact, more large-scale changes in the system, and so forth. However, barring such technicalities, this hypothesis holds (Mailhammer, Restle 30 and Vennemann in press:2). The related Diachronic Maxim extends this to, as the name suggests, change in the language with regard to any given parameter

(8) Synchronic Maxim - A language system will in general not contain a structure on a given parameter without containing those structures constructible within the means of the system that are more preferred in terms of the relevant preference law

(9) Diachronic Maxim - Linguistic change on a given parameter does not affect a language structure as long as there exist structures in the language system that are less preferred in terms of the relevant preference law.

(Mailhammer, Restle and Vennemann in press:2, Vennemann 2000:2)

To determine the relative preference for particular structures, Mailhammer et al

(in press:1) rely on three criteria that are themselves based on Jakobson's earlier insights.

Relative preference is graded on three scales: presence crosslinguistically, sequence of acquisition in first language acquisition, and sequence of loss in aphasia. The more languages have the structure, the earlier the structure is acquired, and the later it's lost in aphasics, the more preferred it is. The criterion of crosslinguistic presence includes the notion of implicational relations in the Greenbergian (1963) sense, such as the presence of voiced stops implying the presence of voiceless ones.

It must be reiterated that these are fundamentally local judgments of preference. 31

That is to say, a preference gradation at one level of linguistic structure can cause changes

that are not preferred or are even dispreferred at another level. This notion of preferences

working at cross purposes accomplishes much the same thing as the Optimality Theory's system of ranked, violable constraints, but not only does not require the generative

assumption that such constraints are an inborn part of the language faculty, but also allows for things like a language using multiple strategies simultaneously to alleviate a dispreferred situation. An example of this would be Pali using both deletion and epenthesis to eliminate word initial clusters; Sanskrit srotas 'stream' becomes Pali sola,

but Sanskrit sneha 'affection' becomes Pali sineha (Mailhammer, Restle and Vennemann

:2, Murray 1982:176).

Mailhammer et al (in press) focus on the structure of the syllable in laying out their preference laws. While a reiteration of all of their work would be excessive here,

the general idea is for now sufficient to indicate that a gradient notion of markedness is not only possible but is in fact being used to good effect here in this thesis.

3.3 Consultation of Existing Reconstructive Work

As far as what resources should be consulted, the top-heavy nature of existing scholarship presents a dilemma. The compact nature of the Hmong-Mien language family, and Qiandong being a subbranch with its place not in question, potentially allows for higher level reconstructions as well as ground-level data to simultaneously influence 32 the lower level reconstruction, a method which Ratliff (2010:21) calls "triangulation".

The typical issue with using high-level information in a triangulation approach, the threat of reconstructing a genetic relationship where none exists by making a reconstruction look like one particular higher-level proto-language or another, isn't present here, as the genetic place of Qiandong is not in question. However, the more existing reconstructions are used to influence choices made in reconstructing PQH, the less potential the PQH reconstruction will have to refine those reconstructions in turn. If concrete language data produces a reconstruction that conflicts with higher-level reconstructions, that may indicate not that the PQH reconstruction is in error but that the higher-level reconstructions are. For instance, Ratliff (2010) only uses a single dialect of Qiandong, the Yanghao dialect, in her PHM reconstruction, as representative of the entire branch. If it turns out that Yanghao is not a typical Qiandongic dialect in some manner, then that has the potential to mislead her whole reconstruction in that regard. Pulling data down from her reconstruction would prevent us from determining this, so it turns out there is, in fact, a reason to hesitate to look at the high-level reconstructive work for arguments.

In this reconstruction I thus take a compromise approach. In general, I attempt to explain the correspondence sets of Qiandongic without taking recourse to looking above to higher level reconstructed protolanguages for information. However, in cases where the Qiandongic data is inconclusive, and broader Hmong-Mien scholarship has a clear and ready answer, it will be taken into account. In particular, this comes into play in this reconstruction in the matter of prenasalized consonants, which are common to other 33

Hmongic branches and have been reconstructed for PHM but do not occur in Qiandong, and in historical tonology, as development of tones in Hmong-Mien is quite well understood and many developments occurred prior to PQH.

3.4 Lexical Diffusion

Returning now to the question of the nature of sound change as discussed in subsection 3.1, two major opposing positions were staked out in the nineteenth century and have been defended ever since. One was the Neogrammarian viewpoint, developed by the eponymous group of linguists, who held that sound change always affected all words with the involved sound at the same time; to rephrase it simply, sounds changed, not words. The other was espoused by the likes of Schuchardt and Gilleron, who argued that each word acted independently. There's been reason over the last hundred years to give credence to both (Phillips 2006).

The Comparative Method, which assumes the Neogrammarian viewpoint as a working principle, has been the backbone of historical phonology for years, and has been quite effective in that regard, as demonstrated by the numerous achievements in linguistic reconstruction. However, seeming exceptions to the Neogrammarian Hypothesis have cropped up. For instance, Sommerfelt (1962:75) found that in some Welsh dialects, the initial voiceless velar fricative's loss affected [xware] 'to play' first, then [xwanen] 'flea', and only after that [xwa:ir] 'sister'. This sort of change affecting some words before 34 others flies in the face of Neogrammarian ideas. The response by those following in the

Neogrammarian intellectual tradition has been to primarily treat these as outliers due to change by analogy with other words or external language or dialect influence. This includes modern generative approaches to phonology such as Optimality Theory, which focuses primarily on the phonological system, which is held separate from the lexicon.

The idea of change occurring on a word-by-word basis in this way was first called

"lexical diffusion" by Wang (1969). In that paper, he outlines four logical possibilities in viewing how a sound change might progress: phonetically and lexically abrupt, phonetically abrupt and lexically gradual, phonetically gradual and lexically abrupt (the

Neogrammarian sort), and phonetically and lexically gradual. The first possibility he regards as impossible, although Janson (1983) has since suggested it may be possible in a very limited sense, when a variant pronunciation from one register is accepted wholesale into another. The third is the Neogrammarian type of change. The second and fourth are the lexically diffused changes, the difference between them being phonetic abruptness or discreteness. While some changes could potentially be gradient, say, vowel height, putting them in the latter category, others are by nature discrete; metathesis, for instance, cannot happen by degrees, and is necessarily of the second type.

The possibility has been raised that "all phonological change starts with lexical diffusion and most ends up Neogrammarian, given time"(Lass 1997:140-141, cf.

Krishnamurti 1998, Andersen 1973:787, Ogura 1995:32, Oliviera 1991:101). However, it's unclear how this could be truly proven, in that it would entail proving that 35

Neogrammarian-style change never occurs, and proving a negative is quite a task.

3.5 Exemplar Theory

In any case, the traditionalist response to the possibility of lexical diffusion has perhaps not adequately quelled concerns about lexical diffusion's threat to the

Neogrammarian paradigm, as more and more examples seem to have popped up since

Chen and Wang's (1975) study of a lexically diffused split in tone categories in Chaozhou

Min Chinese. The approach following in Schuchardt and Gilleron's footsteps, instead of treating these issues as minor anomalies, treats them as having primary importance.

Exemplar models of language, such as Bybee's (2003) and Pierrehumbert's (2001), have a very different view of language, treating the phonological system as a secondary effect arising purely from the interaction of a store of exemplars of perceived and produced language with each other in a connectionist network. Since these models do not posit as a theoretical primitive anything like a phoneme, Neogrammarian across-the-board change is not a direct and natural consequence of such a model, but it does work very well with

Schuchardt's observation that different words change at different rates.

Given both the record of success of Neogrammarianism and the growing body of evidence supporting the existence of lexical diffusion, this paper will take the approach of

Phillips (2006) in treating the idea of a "sound law" as a methodological principle, but not a theoretical one, and presuming that most results will be Neogrammarian in 36

appearance even if the mechanism was lexically diffusive. That is to say, the comparative

method in the traditional sense will still be used here, with appeals to a diffused change in

progress or the like treated as last-resort options, but no assumption is being made that

the changes reconstructed actually happened in a Neogrammarian manner.

Moreover, another methodological issue brought to the forefront by Bybee and

Pierrehumbert's exemplar frameworks is what unit to reconstruct. Again, Bybee

(2003:71) argues that the proper unit by which sound changes should be measured is the

individual muscle movement, a more finely grained unit than even the articulatory

gesture, and that it is at this level that the naturalness of a change should be assessed.

However, this is perhaps beyond the precision of the comparative method, which

is more oriented towards recreating rough categories than intricate phonetic details of

mother language phonologies. As such, this paper will again take a traditionalist

approach in practice. It reconstructs phonemes, although they're assumed to be of the

loose, emergent, derivative type characteristic of exemplar theory rather than as a basic

structural unit of the framework. However, like Phillips, I assume phonetic plausibility is

ultimately determined by activity at the level of the musculature involved. Changes that

are well attested in the literature or phonetically natural even at the surface level will not

need this level of detail, but changes that seem to have no phonetic justification may be

explained by zooming in to the articulatory level.

This focus on the articulatory level of structure enables Bybee (2003:77) to claim,

like Pagliuca & Mowrey (1987) and Mowrey & Pagliuca (1995), that all sound changes 37 with an articulatory basis are at that level reductive or assimilatory in nature. That is to say, they all either involve reduction of the time of a particular muscle action or reduction in its degree. The power in this is that many (and if these claims are right, all) traditional examples of changes going the opposite way, changes that appear to be strengthenings at the phonemic scale, turn out not to be at the motor level, and thus such a strong hypothesis is more easily maintained.

For instance, Pagliuca and Mowrey (1987:462) argue that affrication of voiceless stops in the High German Consonant Shift (e.g. /p/ —»/pf/) is due to an "erosion of stop closure integrity" with acoustic consequences; in other words, a substantive reduction at the level of the muscles forming the stop. Thus, claiming it as a strengthening is misguided.

Another type of change that can be reclassified as reductive in this manner is consonantal epenthesis. In some Spanish verbs, the suffixation of the auxiliary habere 'to have' to several high-frequency verbs caused a stop to be inserted, as seen in (10) below.

(10) venir + he > venire > venre > venire 'I will come'

tener + he > tenere > tenre > tendre 'I will have'

poner + he > ponere > ponre > pondre 'I will put'

(Bybee 2001:73)

The inserted [d] has the same place of articulation as the surrounding consonants 38 for good reason. Rather than viewing this as the insertion of a segment, if the issue is approached at an articulatory level, a different picture emerges. If the muscle movements comprising the nasal element of the [n] are reduced in time without increasing the temporal length of the [r] to compensate for the gap, the sound being articulated at the time between the end of nasality and the onset of the [r] is a [d]. In other words, despite the "insertion" of the consonant segment, no muscle motions were inserted that weren't already there, and in fact one was reduced in time.

Obviously, it would and will take a great amount of painstaking work to determine if such a strong hypothesis of absolute reductivity is feasible. However, from the perspective of delineating a methodology for historical reconstruction, which is the task at hand, maintaining such a strong hypothesis is valuable. If reconstructions such as this can be done without abandoning it, then that is certainly evidence that strengthens it, whereas if it turns out to be necessary to posit changes in the reverse direction, that weakens it. Historical reconstruction can provide a test of this reductive restriction.

It is significant that this idea is only applicable to changes with an articulatory origin, and not all changes need have such an origin. In particular, perceptual motivations, causing active changes by the listener rather than the faithful replication of the speaker's utterance, are also possible. Ohala (1981) details several instances of this in action, and his work will not be repeated here, but the possibility of perceptually motivated changes must be noted. 39

3.6 Summary of Theoretical and Methodological Approach

This third chapter has detailed that this thesis rests on exemplar theory at a foundational level and the non-Neogrammarian implications that arise from it. However, it has also been decided that given the practicalities of historical work and the record of success of Neogrammarian approaches, the thesis uses a largely traditional and

Neogrammarian comparative methodology in practice. This is not an inherent conflict as uniform Neogrammarian results can and perhaps only arise from what were at the time non-Neogrammarian, lexically diffused sound changes. Preference Theory is drawn upon to clarify notions of plausible directions of change, and higher-level reconstructions are consulted only where the Qiand5ng data is intractable in isolation.

4.0 Aspiration

Much of the content and analysis in this thesis, particularly the discussion of fricative development, revolves around aspiration. As such, first it is necessary to discuss how the concept is defined. This is done in subsection 4.1. In subsection 4.2, this is followed up by a discussion of how aspiration relates to voicing. In subsection 4.3, the chapter is summarized. 40

4.1 Defining Aspiration

Over time, there has been several different approaches taken to define aspiration.

Most early definitions revolved around a burst of exhaled air, such as the following:

"The rush of air out of the stopped cavity may be vigorous and puff-like [...] this puff may readily become the distinctive mark of the stop. Stops which have it are called aspirated stops; stops which lack it are called unaspirated stops" (Heffner 1950:120, cf.

Bloomfield 1933:80,99, Gleason 1961:247, Malmberg 1963:42, Peterson and Shoup

1966:103).

Others, noting its effects on VOT, defined it on that basis. In other words, a large time gap between the release of a stop and the onset of vocalic voicing was taken as indicative of aspiration. Lisker and Abramson (1964:387) write, "Aspiration [...] is regarded simply as a large delay in voice onset."(cf. Abercrombie 1967:148, Hockett

1955:40-41). Approaches that regard the aspiration as an approximant segment in and of itself (Kenyon 1962:64, Pike 1943:50) can be considered equivalent to this approach.

However, Kim (1970), in a phonetic study, determined that the width of the glottal opening during the stop is the determining factor behind aspiration. Wide glottal openings produce aspiration, because it takes them more time to close the intervening distance to get in position to begin vocalic voicing. Kim (1970:111) writes, "It follows 41 that no stop is aspirated if it has a small glottal opening, and that a stop cannot be unaspirated if it has a large glottal opening at the time of the oral release."

This definition actually goes against the spirit of some of the previous ones. For instance, Heffner (1950:120)'s claim that "the degree of energy displayed by the puffs depends on the degree of compression achieved during the occlusion" is off the mark, in that it's actually speech organs being farther apart rather than closer together that gives rise to aspiration. This begs the question of where the turbulent nature of [h] comes from.

Kim (1970: 111) claims that the turbulence is in fact generated not at the glottis but at the constriction formed for the following vowel, using as evidence differences in the acoustics of [h] depending on the following vowel.

This he uses to explain several phenomena, as support of his conception of aspiration. The most important one for our purposes is in English; why are voiceless stops after /s/ unaspirated when they are typically aspirated elsewhere? Phonological redundancy is a potential explanation, but Kim (1970:113) provides a phonetic one. In essence, the glottal opening begins as the /s/ begins and is completed too early to provide aspiration. The idea of features allowed Iverson and Salmons (1995:373) to attribute this to a single feature being spread across the whole cluster.

To elaborate on features as they pertain to aspiration and related phenomena and contrasts, Halle & Stevens (1971:51) laid out a three way arrangement of glottal width, with [spread glottis] and [constricted glottis] features, the latter pertaining to globalized consonants and implosives. It must be understood that when features are spoken of here, 42 they are spoken of as markers of a pattern, in this case a pattern of aspiration. However, this is by no means intended to be interpreted as anything but a property emergent from the mass of exemplars in the speaker's memory.

4.2 Aspiration and Voice

Iverson and Salmons (1995:383) use the features [spread glottis], [constricted glottis] and [voice] to establish a typology of languages in terms of those features and their implementations, which is less straightforward than it seems at first glance. In

English, because the [spread glottis] feature, as discussed above, can be spread in the case of clusters that do not produce phonetic aspirates, all English voiceless stops can be said to have that [spread glottis] feature. Voiced ones take a [voice] feature instead.

There is a connection then made here between voicelessness and aspiration.

While this example makes that connection with obstruents, sonorants have a stronger one yet. Cho (1991) in fact pulls together evidence that for so-called voiceless sonorants, the important contrast is not a lack of voice, but the presence of [spread glottis]. Iverson and

Salmons (1995) provide empirical evidence for this from English consonant clusters. In clusters with liquids after voiceless obstruents in English, as can be seen in (11), the liquids are voiceless. Much like in the clusters starting with /s/ in Kim (1970), rather than needing to posit a phonological rule or its framework-appropriate equivalent for this, if one assumes that the feature [spread glottis] attaches to the whole cluster and causes the 43 perceived devoicing, there is no empirical difficulty. The stop, not bearing its own

[spread glottis] feature to keep the glottis spread during production of the stop, coincides in timing with the closure of the glottis. In contrast, attempting to explain the stop's voicelessness by rule would be difficult, as such a rule would have to distinguish between voicelessness in initial stops, which would otherwise result in perceived aspiration, from its occurrence in clusters where aspiration does not appear.

(11) plan [plan] shrimp [fjimp] crow [Iqo] sneeze [sniz] slip [slip] fleet [flit] (Iverson and Salmons 1995:373)

Another argument for aspiration and not voice as the pivotal feature in "voiceless" sonorants comes from Burmese (Mester and Ito 1989:280). Burmese contrasts three obstruent series (voiced, voiceless and voiceless aspirated) and two sonorant series

(voiced and voiceless aspirated). The voiceless aspirate series of sonorants, written hi, hm, hn, hug, hny, and hw, has been described by Okell (1969:9) as the phonetically aspirated counterparts of written I, m, n, ng, ny, and w, "with the breath expelled quietly through the nose (through the mouth for hi and hw) before voicing begins..." Also, voiceless aspirated sonorants can be generated when an optional elision process 44

eliminates the phoneme in between a voiced sonorant and an /h/, as demonstrated in (12)

on the following page. This is analogous to a phenomenon noted by Kim (1970:113) in

Korean, amongst other languages, in which an unaspirated stop and an adjacent /h/

combine into an aspirated stop, as a coarticulation effect; the wide opening of the glottis

inherent in the /h/, simultaneously pronounced with the stop, creates aspiration. As a side

note, a similar process is thought to have taken place in the development of Indo-

European, as per the well known Proto-Indo-European laryngeal theory (Lindeman

1987).

(12) /mahou?/—>[mhou?] 'be not true' (Mester and Ito 1989:280)

There is also a rule in Burmese in which morpheme-initial consonants voice when

word medial due to morpheme concatenation (1989:280). (13) below contains some examples. This is a productive rule, and only fails to take effect when a glottal stop ends

the previous syllable. Interestingly, while plain and voiceless aspirated obstruents become voiced under this rule, aspirated sonorants are unaffected. Mester and Ito

(1989:280) draw the conclusion that only obstruents can be marked for [voice]. 45

(13) [poup] 'can' [shi+boup] 'oil can' [the] 'inside' [eijide] 'in the house [nhi?] 'year' [kou+nhi?], *[kou+ni?] 'nine years'

More evidence comes from Klamath (Clements 1985:234). Phonological

processes existing in that language have the effects listed in (14) below.

(14) /nl/—*[11] /nl/—>[1] /nP/-+[P] /11/->[|] /11V—>[P]

In the case of Klamath, the theoretical benefit of assuming the voiceless [1] to be aspirated is that then all consonants, obstruent and sonorant alike, in Klamath can be classified into three series, plain, globalized and aspirated. Moreover, this also would allow the rules in (14) to be simplified to two, a lateralization of the /n/ and the spread of the [spread glottis] or [constricted glottis] to the whole cluster (1985:234).

Finally, a relevant case occurs in Icelandic. In the standard Icelandic dialect,

/l,m,n/ devoice prior to /p,t,k/, which concurrently deaspirate (Thrainsson 1978:39).

Other dialects have the two phenomena to varying degrees, but they are inextricably tied; 46 no dialect, however, has one phenomenon without the other (Cho 1991, Thrainsson

1978:39), Table 3 below shows examples of this. Phonetic studies indicate that the glottis is spread during the articulation of the sonorants, but has begun to close by the time the stop occurs (Petursson 1976:198, Thrainsson 1978:40). This is exactly what would be predicted by Iverson & Salmons (1995:373)'s idea of a single [spread glottis] feature dealing with both devoicing and aspiration spreading across a cluster. This feature spreading manifests in the stop coinciding in timing with the closure of the glottis.

Table 3: Icelandic Sonorant Devoicing and Stop Deaspiration

Written Gloss Standard/ Dialect Ungrammatical forms devoiced without ; dialect devoicing ulpa 'coat' Mpa] [ulpha] i *[ulpha], *[ulpa]

heimta 'demand' [heiipta] [heimtha] *[heimtha], *[heimta] vanta 'lack' [vanta] [vantha] *[vantha], *[vanta] vinka 'wave' [vigkha] h [virjka] *[vigk a], *[viqka] (Thrainsson 1978:39)

Given this array of evidence, those sonorants listed as voiceless in Ma & Tai

(1956) will in this thesis be given as aspirated instead. It is, in effect, the same thing phonologically, and allows them to be written in the same fashion as Purnell (1970).

Another concern regarding aspiration is the representation of an aspiration distinction in fricatives. Halle and Stevens (1971:55) treat voiceless fricatives as generally [-spread glottis], and [+spread glottis] in the case of aspiration, and this view 47 was adopted in practice by others (eg. Catford 1988, Maddieson 1984a). However, later phonetic work, namely Kingston (1990) and Stevens (1991), indicates that voiceless fricatives are produced with the glottis spread. To account for this, Halle (1995) reverses course and treats all fricatives as [+spread]. However, under such a scheme another method would be required to contrast the aspirated and unaspirated fricatives present in the data in this thesis as well as in many other languages (Jacques 2011). It also ignores another phonetic result, that voiced fricatives are tied to a narrowed glottis (Catford

1977:112).

Vaux (1998) argues with evidence from a variety of languages, including

Armenian, Sanskrit, Greek, Spanish, and Thai, that in general, voiceless fricatives are

[+spread glottis] and voiced ones are [-spread glottis]. However, to avoid the problem with languages that maintain aspiration distinctions in fricatives, he claims that this is merely the least marked arrangement for fricative systems, not a crosslinguistically universal specification. He writes, "It is perfectly possible for languages to contrast

[spread glottis] values in voiceless fricatives allophonically (as in English) or phonemically (as in Burmese)" (Vaux 1998:508).

This begs the question of what exactly is being recorded when a fieldworker transcribes an aspirated fricative. Phonetic studies are lacking regarding the exact phonetic qualities of the aspirated fricatives in Southeast Asia, but presumably, along the lines of Heffner's (1950) definition of aspiration at the beginning of subsection 4.1, Ma &

Tai (1956) and the sources used by Purnell (1970) were hearing increased airflow relative 48 to the series of fricatives they marked as unaspirated, potentially from an increased degree of glottal compression. In languages where unaspirated fricatives are [-spread glottis], they would provide a baseline degree of airflow and compression from which

[+spread glottis] aspirated fricatives can be contrasted. More research in this area would be highly informative, but interpreting the transcriptions as having stronger "puffs" of air, in the sense Heffner (1950) used the term, is adequate in the interim for reconstructive purposes.

While this might not be the most palatable solution were we trying to establish a phonological specification for fricatives as a linguistic universal, remember that working within the exemplarist framework, as we are in this paper, these features are only a convenient way of tracking linguistic patterns, not a theoretical primitive of any kind. As such, the notion that all languages should have the same feature specification arrangement in their fricative system can be discarded without consequences.

4.3 Summary Of Aspiration Discussion

Aspiration is thus tied, phonetically, to a spread glottis. Related phonetic data suggests that as a result, "voiceless" sonorants can be considered to be aspirated. Also, this definition of aspiration has led to the notion of a [spread glottis] feature in the phonology. However, no featural arrangement has proved adequate crosslinguistically, particularly with regard to aspiration distinctions in fricatives. Since such distinctions 49 play such a prominent role in Qiandong phonologies, this thesis in practice follows

Vaux's (1998) assessment that in languages that contrast aspiration in spirants, a distinction in [spread glottis] exists, but in languages that do not, all fricatives are

[+spread glottis],

5.0 Plosives

Here begins the reconstruction itself. In this chapter, the development of the plosives is detailed. The Qiandong languages represented in the data seem to have five places of articulation for plosives, with a two-way aspiration distinction. This is nothing unusual, particularly for Hmongic, which has attested reflexes in all of these places of articulation and widespread aspiration distinctions in stops. At least three of the series, the set {ph, th, kh}, are continuous without interruption and can thus be reconstructed as- is, but the other seven in the Ma & Tai (1956) data require further explanation. Tables 4a and 4b on the next two pages show the plosive reflexes in Qiandong. 50

Table 4a: Plosives in Ma & Tai Dialect Series Series Series Series Series Series Series Series Series Series 1 2 3 4 5 6 7 8 9 10 Congjiang P t c k q PH th ch kh qh Huangli p/ph t/t'1 c/ch k/kh q P" th ch kh qh Jiaba P t c k q PH th ch kh qh Jlnping P t t? k q P" th t9h kh qh Jingxian P t t? k q P'S th t?h kh qh Jlnzhong P t c k q ph t" ch kh qh Jiuzhou P t c k q P11 th ch kh qh Paitmg P T c k q ph th ch k" qh Sandu P t c k k ph th ch kh kh Sansui P t c k q p'1 th ch kh qh Taigongzhai P t c k q P'1 th ch kh qh Taiyong P t c k q PH th ch kh qh Wucha P 't c k q ph t" ch kh q" Wuluo P t c k q ph th ch kh qh XInqiao P t c k q ph th ch kh qh Xuanwei P c k q ph th ch kh qh Yong'an p/ph t/th c/ch k/kh q ph th ch kh qh Zhenyuan p t c k q ph th ch k" qh Zhouxi P'Ph 1 t/th c/ch k/kh q PH th ch kh qh (Ma & Tai 1956:43-45) 51

Table 4b: Plosives in Purnell

Dialects Ser. 1 Ser. 2 Ser. 3 Ser. 4 Ser. 5 Ser. 6 Ser. 7 Ser. 8 Ser. 9 Ser. Ser. 10 11 Gaotongzhai ptt--k--th- Kaitang p t

ShTdongkou p t

Yanghao p t kh

Zhenfeng p j t (Purnell 1970:46-49)

Subsection 5.1 discusses uvular and velar place changes, subsection 5.2 discusses a tone-related aspiration merger in tone 8 stops, 5.3 discusses affrication in palatal series,

5.4 elaborates from a typological perspective on the addition of prenasalized onsets to the reconstruction, and section 5.5 visually summarizes the chapter.

5.1 Uvulars and Velars

First, one notable consonantal development that must be remarked upon here is with regards to uvulars and velars. Otherwise uninterrupted correspondence sets of aspirated and unaspirated uvular stops are velar in Sandu, as seen above. Table 5 on the next page shows some related correspondences from Ma & Tai (1956:38). 52

Table 5: Uvular-Velar Stop Correspondences

Gloss Taigongzhai Sandu PQH

'dung' [qa3] [ka3]

h 'tie up' [qhalj [khal] *q

'chicken' [qalj [k8l] *q

The important thing about this set is that 'chicken' has been noted in Rati iff

(2010:16) as a loan from Chinese, as shown in (15) below. Old Chinese (OC), Middle

Chinese (MC), and Mandarin forms are given for comparison; the Mandarin form is given in pinyin. It should be noted that the tone of the Proto-Hmongic (PH) reflex is listed as tone A, as the tone split based on initial voicing, discussed in subsection 1.3, had not yet occurred to split the tone into the contemporary A1 and A2 (1 and 2) categories.

Finally, the exact quality of the PHM initial involved is not entirely settled, and thus has been capitalized to indicate this.

(15) OC *kfe > MC kej > Man.yT 'chicken'

PHM *Kai > PH *qeA

As can be seen, this was originally velar, at least in Chinese. Ratliff (2010:16, cf.

Solnit 1996:13-14) argues that some, but not all, modern uvulars, descend from PHM velars, while others descend directly from PHM uvulars, in a case of what Hoenigswald 53

(1960) termed a primary split, as discussed in subsection 3.1. Ratliff (2010:16) reconstructs the remaining velars as having had their retraction blocked by an intervening back rounded vowel or onglide. There is areal support for this analysis as well; Matisoff

(2003:20) writes, "[p]ostvelars are generally secondary deveolopments of the TB [Tibeto-

Burman] *velar series, as in Black Lahu, where they regularly descend from simple

*velars that are not followed by a glide".

However, 'tie up' in Table 5 appears to be natively Hmong-Mien and descended from an uvular, Ratliff (2010:270)'s reconstruction of it for PHM being *qheA. Table 5 indicates that its correspondences across Qiandong are the same as those of the derived uvular in 'chicken'. This indicates that whatever has happened in Qiandong did not draw a distinction between the two, and the velar retraction process can be presumed to have been completed before PQH.

What seems to be happening, then, is merely a merger in Sandu, where uvulars of both origins, regardless of environments, have fronted to velars, while in all other

Qiandong dialects they remain uvular. The motivation for this merger remains unclear, but perceptual confusion between the two places of articulation may be a factor.

5.2 Disappearance of the Aspiration Distinction in Tone D2

Aspiration appears to be affected by tone in some Qiandong varieties. In the tone

D2 in the Yong'an, Zhouxl and Huangli dialects, the tone marked as 8 in the tables, as 54 described in subsection 1.3, the reflexes of the series that are elsewhere unaspirated plosives and affricates are aspirated; the aspiration distinction is lost in that tone with regard to plosives and affricates. Table 6 below shows this, with the Taigongzhai reflexes provided as a point of reference. Note that in ZhouxT tone D2 has merged with tone D1.

Table 6: Aspiration Distinction Loss In Stops

Huangli Taigongzhai Yong'an Zhouxi

bucket pheg8 pu8 phu8 pi,u7/8

cat ph£8 pei8 ph£8 pii£7/8

piece of cloth - tsag8 - tshei7/8

8 pocket tha8 te8 tha8 tha7/

pot khai]8 kaq8 khai]8 khe7/8

see Pheij8 pu8 phu8 jphu7/8

8 sour liquid qhai]8 qaq8 - qh_j7/

ten t9hau8 t£08 t9hu8 t9hu"^8 (Ma & Tai 1956:38)

The simplest solution would of course be to reconstruct a merger here in tone D2 in the three dialects. The alternative would be a split with no clear conditioning factor in all other tones and dialects, a much, much larger set of environments. Such a merger can be motivated by the qualities of the tone itself. In particular, this merger is suggestive of the effects of phonation in tone. 55

Elsewhere in Hmongic, phonation is known to be intertwined with tone; for instance, in both Green Mong and White Hmong, there are "falling creaky" and "falling breathy" tones, D2 and a merged B2/C2, respectively (Mortensen 2004, Golston & Yang

2001). In fact, this phonation has been analyzed as the primary acoustic cue in White

Hmong (Yang 2000) and Niederer (1998:236) mentions combined pitch/phonation cues in tones as a characteristic of the Hmong-Mien family more broadly. Thus it is quite possible that phonation is present in the tone here as well.

Green Mong's "falling breathy" tone never occurs concurrently with aspirated onsets (Mortensen 2004:4). One very striking possibility is that the cue of breathiness, from the spread glottis, that is required by the tone, interferes with the perception and indeed the production of a distinction in consonantal aspiration, as it relies on the same cue of breathiness produced by a spread glottis. If this is indeed the case, then the unaspirated onsets in Table 3 would all sound aspirated.

However, Ma & Tai (1956) makes no mention whatsoever of phonation in its discussion of tone values. Hmong-Mien fieldwork, reconstruction and internal typology was in its infancy in 1956, so it is quite possible they simply did not know to look for it, or lumped breathy phonation in with aspiration.

In any case, even if the omission of phonation details was not a lapse and phonation is currently absent, phonation effects at a prior state of historical development might produce a collapse of the distinction that remains synchronically. This seems likely at least in the case of ZhouxT, as tone D1 words with the affected onsets show no 56 change, and thus this change presumably precedes the tone merger. As such, until more is known about tonal developments in Hmong-Mien, it suffices to say that the effects of breathy phonation in tone D2 are probably responsible for what is seen here.

5.3 Palatal Stops and Affricates

In the palatal stop series, Jinping and Jingxian have affricate reflexes. Changes from stop to affricate are common and has been argued to be a type of lenition (Lass

1984, Honeybone 2002), and could be posited here on that basis. Another potential motivation for positing an affrication is maximization of perceptual contrast (see Hock

1991, Martinet 1955), often termed polarization or distantiation; as there is a large number of stops in the PQH system, the affrication may be increasing perceptual contrast by moving the palatal series out of that perceptual space. Unfortunately, as can be seen in

Table 12a in Chapter 7, predominantly palatal affricate series also have a few stop reflexes, and if stops are reconstructed here, that would imply that those series, following majority rule, has changed in precisely the opposite direction.

The picture is dramatically simplified, however, if the prenasalized consonants of other Hmongic subgroups are assumed at PQH level. This allows for the two series to be differentiated by prenasalization, thus avoiding the dilemma of both types of series seeming to warrant the same reconstructed stop, and allowing an affrication to be reconstructed for both series. Additionally, prenasalization has already been used in 57

Wang's (1979) work on the origin of Qiandong aspirated palatal and alveolar fricatives,

the conclusions of which are discussed and adopted in Chapter 7. Table 7a below shows

examples of the predominantly plosive series with affricate reflexes from Jingxian, while

7b shows examples of the largely affricate series with stop reflexes from Sansui, both

with the reconstructed PQH onset. The former is reconstructed as a prenasalized *nc and

*nch, while the other affricate series is reconstructed as *c and *ch.

Table 7a: Prenasalized Stop Series

Gloss Jingxian T aigdngzhai PQH

1 h 'change' [tghu3] lcl!o3] *nc

'chopsticks' [t?o6] [cou^] *nc

'crab' [t9au 1 ] [CO1] *nc

; h 'shave' [t?hi5] [chi^] *nc

Table 7b: Non-prenasalized Stop Series

Gloss Sansui Taigongzhai PQH

h cloth [c'W] [t9hul] *c

'medicine' [ca1] [t^a1] *c

h 'money' [che^] [t9he^] *c

'succeed' [C£2] [t9aq2] *c

h 'wipe' [che2] [t9he2] *c (Ma & Tai 1956:37) 58

5.4 Origin and Plausibility of Prenasalized Onsets

Prenasalized consonants are "a hallmark feature of the phonological inventories of most Hmongic languages"(Ratliff 2010:207) and are reconstructed for PHM (Ratliff

2010), but are not present at all in modern Qiand5ng. In fact, along with aspirated fricatives, the absence of such prenasalized initials was a feature used by Purnell

(1970:13) to delineate the Qiandong subgroup in the first place. Yet, as was mentioned in section 5.3, positing prenasalized initials in PQH is necessary to explain the development of the aspirated fricatives and the current synchronic distribution of stops and affricates.

A note is warranted, then, about the origin of these sounds, if they are to be posited with no surviving reflexes. Ratliff (2010:207), reconstructing them for PHM, posits that they originate from elements fused to the onset that have otherwise disappeared; in particular, she argues for considering them either residue from a preceding disyllable or a fragment of a prefix. The disyllabic analysis is hampered by the fact that the only disyllabic forms reconstructed in Ratliff (2010)'s PHM with a nasal in the first onset are in instances where later forms cause the first nasal to replace another nasal in the surviving syllable's onset, the three relevant instances being *m-nok 'bird',

*n-mej 'to have' and *n-m»q 'to go'. However, the form for bird is tantalizingly close to

Proto-Malayo-Polynesian *manuk 'bird'; while, even if this is a loan, it is unknown whether it was borrowed in disyllabic form or reduced in the borrowing process, it is a distinct possibility. 59

Outside of the forms reconstructed with nasals, there is some evidence for the preceding syllable hypothesis. For instance, the variety in onset clusters in Hmong-Mien languages, including two different patterns of correspondence implying nasality in the initial, leads Ratliff (2010:12) to reconstruct both "tight" and "loose" consonant clusters in PHM, the latter in her words "not being clearly distinguishable from a disyllabic structure" (2010:206). Sagart (1999:13) makes a similar distinction of closely and loosely attached preinitial elements in his reconstruction of Old Chinese, while Proto-Tai

(Pittayapom 2009) and Proto-Tibeto-Burman (Matisoff 2000) have been reconstructed as sesquisyllabic, and Mon-Khmer (Thomas 1992) has widespread "syllable-and-a-half' structures, so to think that there could be a preceding syllable or demi-syllable is quite sensible from an areal viewpoint.

As for the possibility of the prenasalization coming from a prefixial remnant, prenasalization only appears morphological in origin in a handful of Chinese loanword pairs. However, these pairs are somewhat intriguing. First, Downer (1973:14-16) notes two pairs of words in modern Mien, both loans from Chinese, shown in (16) below, in which prenasalization, represented with a voiced reflex in the second member of each pair, seems to be morphologically active.

(16) /tslie?7/ 'to pull down/apart' vs /dze?7/ 'to be cracked'

/khoi1/ 'to open (transitive)' vs /goi7/ 'to open (intransitive)' 60

As can be seen, the prenasalization seems to be relevant for a transitivity distinction. Downer (1973) believes this is indicative of a prefix on the intransitives at the Proto-Mienic level, and the fact that these are Chinese loans is to him a sign that the prefixation was productive at the time of borrowing. On the other hand, Ratliff

(2010:208) writes, "It seems more likely that this morphological contrast was borrowed from Chinese along with the words themselves [...] elsewhere, prenasalization seems to be phonological". The only potential morphological contrasts based on prenasalization located by Ratliff (2010:208) are listed in (17) below. The semantic distinctions in each pair seem to be separate, and thus under a morphological analysis would represent separate morphemes being preserved as prenasalization.

(17) PHM *GaX 'low/short' vs Proto-Hmongic *NGa(Y) & Proto-Mienic *ya 'to descend'

PHM *tshjieq 'new' vs PHM *ntshjiei3 'clear'

In any case, while sufficient evidence is lacking to solidly support either analysis of the origin of Hmong-Mien prenasalization, it suffices for the purposes of this work to note that both are plausible, and as such positing prenasalized consonants in PQH does not merely shift an inexplicable problem up the family tree. The question of how they arose in Hmong-Mien may or may not remain a mystery, but their existence does not pose any fundamental problems for our analysis or our framework for understanding linguistic change.

5.5 Summary Of PQH Plosive Inventory And Changes

Below in Figure 4 is the set of all consonants reconstructed in this chapter for

PQH. Also, in (18) below is a list of changes posited in this chapter.

Figure 4: Reconstructed PQH Plosives

*P *p!l *t *k *kh *q *qh *nc *nch

(18) PQH {*ph, *th, *kh} no changes

Palatal Denasalization and Affrication - PQH *nc(h) > t9(h) in Jinping and

Jingxian, c(h) in other dialects

Palatal Affrication - PQH *c(h) > t

Tone 8/D2 Stop Aspiration Merger - PQH {*p, *t, *k} > {ph, th, kh} in tone 8 in

Yong'an, ZhouxT and Huangli

6.0 Nasals

In general, in Qiandong there are three places of articulation where nasals are contrastive in aspiration, namely bilabials, alveolars and palatals, while there is also a velar series in which aspiration is not contrastive. However, there are multiple distinct anomalies. Tables 8a on this page and 8b on the next provide the correspondence sets from Ma & Tai (1956) and Purnell (1970), respectively.

Table 8a: Nasals in Ma & Tai

Dialect Series 1 Series 2 Series 3 Series 4 Series 5 Series 6 Series 7

Congjiang in n/nz 1 m n D Huangli m n/z 1 h h h 0 Jiaba m n Ji mh nh 0 j Jinping m n Ji m n J> 0 Jingxian m n Ji m n J» 9 JTnzhong m n Ji mh nh 0 Jiiizhou m n Ji mh nh ji" q Paiting m n/nz 1 m n q

Sandu m n/nz 1 m n J" 0 Sansui m n J* m n Ji D Taigongzhai m n J» mh nh Jih 0 Taiyong m n/z 1 m n .p D Wucha m n Ji mh nh D ; Wuluo m n/nz 1 mh nh h L ______- J* D Xlnqiao m n iji mh nh Jih 0 Xuanwei m n jji mh nh

Yong'an m n mh nh Jih 0 Zhenyuan m n Ji mh nh ji" 0 Zhouxl m n/nz mh nh jih JJ (Ma & Tai 1956: 41,44) 63

Table 8b: Nasals in Purnell

Dialects Ser. 1 Ser. 2 Ser. 3 Ser. 4 Ser. 5 Ser. 6 Ser. 7 Ser. 8 Ser. 9 Gaotongzhai - m n Kaitang m m" n

Shidongkou m mh n n nh

n Yanghao m m n Ji K

Zhenfeng | m m I m n n'1 J> .n ,nh (Purnell 1970:50-51)

Subsection 6.1 discusses a merger in aspiration in nasals with the distinction, and a lenition to h of those same nasals, 6.2 details a merger of velar nasals with palatals in

Xuanwei, 6.3 notes and leaves for future research a problem with nasal-fricative sequences, 6.4 posits a lateralization process, and 6.5 summarizes.

6.1 Nasal Aspiration Contrast

Aspiration contrasts in nasals are common to most Hmongic varieties, and much of Qiandong is no exception. In some of the languages of the branch, aspirated and unaspirated nasals are contrastive, while in others, they are not, and in Huangli the aspirated nasal is instead a /h/, examples of which are shown in Table 9 on the next page. 64

Table 9: Nasal Aspiration Merger

Gloss HuangH Taigongzhai Taiyong PQH

'heavy' [hou3] [PhoiJ3| [jioy3] *Ph

'many' [no5] [ne^] [no 5] *n

h 'sun' [ha1] [nha'] [na1] *n

h 'tooth' [hi3] [mhi3] [mi3] *m (Ma&Tai 1956:35,47)

It would seem the simplest way to treat this would be as a case of merger in those languages without the contrast, and as simple lenition in HuanglL Indeed, the merger analysis is corroborated by the existence of an aspiration contrast in many other Hmongic varieties, and given the similarity between an aspirated sonorant and [h], the lenition is also quite an intuitive one. As such, both aspirated and unaspirated nasal series are reconstructed here for PQH in palatal, alveolar and bilabial position. Purnell (1970) lists no aspirated velar nasal series anywhere in Hmong-Mien, so the velar nasal lacking the contrast can be explained simply by it never having had an aspiration contrast to preserve from previous stages of evolution.

6.2 Palatal-Velar Nasal Merger in Xuanwei

The otherwise exceptionless velar nasal series in Table 8a is palatal in Xuanwei; this is indicated in Table 10 below. 65

Table 10: Xuanwei Nasal Merger

Gloss Taigongzhai Xuanwei PQH

,|azy' [oi4] [pi4i

; 2 2 'slanting' [0a ] [pi ] (Ma&Tai 1956:37)

The best solution here is a merger; interpreting these correspondences as a split would not only force the reconstruction to posit an environment, it would conflict in a most dramatic way with the Majority Rule principle. A merger, in contrast, has neither problem, and moreover it could be explained as a collapsing place distinction in a crowded perceptual space, reducing the number of distinct nasal onsets from 7 to 6.

Notably, such a merger would have to have happened after the aspiration contrast in nasals developed, or a similar contrast would have been generated in the former velars.

This is in line with what would be expected if, as Purnell's (1970:50) correspondences suggest, the aspiration distinction in non-velar nasals goes back to PHM.

6.3 Nasal-Fricative Sequences

Nasal-fricative sequences are unfortunately a puzzle with too little data to solve at this point. Alveolar and palatal nasals alternate across dialects inconsistently and without clear conditioning factors with fricatives ([z] and [j]) and sequences of nasal and fricative 66

([nz] and [pj]). No synchronic systematicity is apparent. A sample of the unpredictable correspondences is given below in Table 11.

Table 11: Unpredictable Nasal-Fricative Sequences

Gloss Taigongzhai Wuluo Yong'an Zhouxi

'boat' [paq2] [pau2] 1 [pag2] [pjag2]

'grass' [nag2] [nzat]2] ! [zaq2] [nzat]2]

'green' [nau2] [nzau2] ; [zo2] [nzo2]

'man' [nei2] [na2] |[ne2] [ne2]

'many' [ne^] [no^] [no-*] [no^] (1956:39)

Ma & Tai (1956:39) write:

"An initial in one place may correspond to two or more initials in another place, apparently without statable conditioning factors [...] Some words with [n-] and [p-] in

Taigongzhai correspond to [n-] and [p-] whereas others correspond to [nz-] and [pj-] in

Wuluo, ZhouxT, Pairing, Congjiang and Sandu; some correspond to [n-] and [p-] whereas others correspond to [z] in Yong'an, Huangll, and Taiyong."

Any potential answer to this puzzle, be it a compound, prenasalized unit of some kind, or something else entirely, will need to find conditioning factors. Ideally, further 67

work on Hmong-Mien, particularly on Qiandong rimes, will make those conditioning

factors more apparent. In the meantime, the anomalous reflexes will have no

representation on the charts of reconstructed sounds, and those words that are

consistently nasals in all dialects are considered to be representative of the appropriate

nasal phoneme.

6.4 [I)-(ji] Alternation

The final issue with nasals is the alternation of [1] and [ji] present across dialects in Table 8a's Series 3. Positing *1 as the protoform poses many problems, not the least of which is that the dialects with [ji] also have an [1] phoneme, which would indicate a split with no clear conditioning factors. Moreover, it would go against majority rule,

would unnecessarily complicate matters of lateral development, and would disrupt the symmetry of aspirated and unaspirated nasals, forcing the thesis to posit the less preferred

*jih for Series 6 but not the more preferred *p for Series 3 as PQH protoforms. If *ji is

the protoform for Series 3, however, this becomes analyzable as a simple merger, and again, distantiation could be a motivation for such a merger, as a system with seven nasal onsets has a very tightly packed perceptual space; lateralization of *p reduces this to 6.

However, why *ji would lateralize to [1] and not [l>] is unclear. In any case, the best solution here is clearly to continue to posit *ji as the protoform and assume merger and lateralization. Unfortunately this alternation is strangely given without examples in the 68 commentary in Ma & Tai (1956), so no table will be placed here.

6.5 Summary of PQH Nasal Inventory and Changes

Figure 5 below is a visual summary of PQH onsets posited in this chapter. (19), also below, summarizes sound changes posited.

Figure 5: Summary of PQH Nasal Inventory and Changes

h I *mh *n i

(19) Nasal Aspiration Merger - PQH [+nasal] > [-spread glottis] in Congjiang, JTnping,

Jingxian, Paiting, Sandu, Sansui, and Taiyong

Huangll Nonback Nasal Lenition - PQH [+nasal, -back] > h in Huangll

Xuanwei Velar Nasal Fronting - PQH *ij > ji in Xuanwei

Nasal Lateralization - PQH *ji > 1 in Congjiang, Huangll, Paiting, Sandu,

Taiyong, and Wuluo

7.0 Non-Lateral Fricatives and Affricates

This chapter, the largest in the reconstruction, concerns itself with fricatives and affricates other than laterals. Tables 12a and 12b on the next three pages show the fricative and affricate dominated correspondence series in Ma & Tai (1956) and Purnell

(1970), respectively.

Table 12a: Fricatives and Affricates in Ma & Tai

Dialect Ser. 1 Ser. 2 Ser. 3 Ser. 4 Ser. 5 Ser. 6 Ser. 7 Ser. 8 Congjiang v ts it? i tsh t?h

Huangll ts tg/t?h tsh t?h f

Jiaba It? t?h t?h f

Jlnpfng ts t? tsh t?h f ts

Jingxian v J ts t? tsh t?h if

Jinzhong |j ! ts t? tsh t?h ! f

Jiuzhou ij ft? t? t9h t?h s

Paiting ts tsh t?'1 f

Sandu ts t? tsh t?h f

Sansui ts tsh 1 ch ts

Taigongzhai |v ts t? tsh I t?'1 ; s

Taiyong v ts tf tsh I t?1'

Wucha J t? t?h ! f/f1 1 s/sh

Wuluo ts t? |tsh ;t?h 70

XTnqiao V j *9 t9 t9h t9h f

Xuanwei V j ts t9 tsh t9h f

Yong'an V j ts/tsh t9/t9h tsh t9h f

h h Zhenyuan V i s *9 s t9h f/f' s/s

ZhouxT V j 1 t9/t9h tsh t9h f/f' s/sh

Dialect Ser. 9 Ser. 10 Ser. 11 Ser. 12 Ser. 13 Ser. 14 Ser

Congjiang h 9 f s 9 Y x Huangll h 9 f1 sh 9h Y x Jiaba h 9 f1 sh ?/kh Y x . JTnping h 9 f s 9 i X

Jingxian h 9 f s ? j/v X

__ __ JTnzhong h 9 P1 sh 9h Y X Jiuzhou h 9 f> sh 9h Y X Paiting h 9 f s 9 Y X Sandu h 9 f sh 9h Y k"

Sansui h 9 f s 9 i X

Taigongzhai h 9 P sh 9h Y X Taiyong h 9 f tsh t9h Y X Wucha h 9 f1 sh 9h Y X Wuluo h 9 f1 sh 9h Y X XTnqiao h 9 f1 sh 9h Y X Xuanwei h 9 f> sh 9h Y X Yong'an h 9 F> sh 9h Y X 71

Zhenyuan h kh

Zhouxi h 9/9'' y/w (Ma & Tai 1956)

Table 12b: Fricatives and Affricates in Purnell

Dialect ' Ser. 1 Ser. 2 : Ser. 3 | Ser. 4 Ser. 5 Ser. 6 Ser. 7 Ser. 8 Ser. 9 Gaotongzhai - - . s - - - t? - h Kaitang - - sh tc

Shidongkou K *9

Yanghao j s tf t9h

Zhenfeng t? t5" Dialect ! Ser. 10 Ser. 11 Ser. 12 Ser. 13 'Ser. 14 Ser. 15 Ser. 16 Ser. 171 Gaotongzhai - Kaitang

Shidongkou i v

; Yanghao

| Zhenfeng f ; sh |9 19 (Purnell 1970: 46-49)

[v], [h], and Q] in Table 12a have exceptionless series and as such are reconstructed as-is. The remainder of the chapter is laid out as follows: subsection 7.1

1 Purnell (1970:49) gives Kaitang and Yanghao's velar fricatives as aspirated, but given that there are no unaspirated counterparts in either dialect, and the phonetic indistinctness of velar aspiration, they are ignored here 72 discusses a chain shift in Qiandong that both created the sibilant aspirated fricatives and reduced the PHM prenasalized onsets, 7.2 expounds an analysis of the development of aspirated labiodental fricatives, 7.3 explains mergers in aspiration in fricatives, 7.4 explains velar stop reflexes in velar and aspirated palatal fricative series using a perceptual rationale, 7.5 discusses the reconstruction of Series 14 from Table 12a and its mix of reflexes, and 7.6 concludes and summarizes.

7.1 Qiandong Chain Shift

The remaining affricates and the alveolar and palatal fricatives have a relatively complicated set of reflexes. However, there has been work done on these already, that simplifies the task of their reconstruction tremendously. Specifically, Wang's (1979)

PHM analyzes them as the result of a series of changes, a chain shift causing the loss of the prenasalized onsets present elsewhere in Hmongic. This chain also results in aspirated palatal and alveolar fricatives at the PQH level. Being able to not only explain these two essential Qiandongic traits but tie them to each other is highly compelling and as such his chain shift, displayed below with regard to his Proto-Hmong-Mien and the

Yanghao dialect, has been adopted here. 73

(21) nts-> s-> sh-

nt§- > G- > G'1-

ntsh-, -tsh > sh-

nt§h- > eh-

This chain shift, when decomposed into two parts, has analogues elsewhere. In one part, the removal of the fricative [s] from the system causes a cascading reduction of the prenasalized affricate onsets to fill the gap, preserving their own aspiration distinctions as they do so; this chain shift has a Burmese analogue. The second aspect of the Qiandong chain shift is that removal of [s] occurs, unlike in Burmese, by way of the unaspirated [s] aspirating; a similar phenomenon has, however, taken place in Shan, a

Tai-Kadai language spoken in eastern Burma. These two subsets of the changes in

Qiandong are laid out in (22) and (23) respectively, below. Note that the sets overlap; the reduction of unaspirated prenasalized onsets to unaspirated fricatives is part of both the

Burmese-type and Shan-type changes, and as such is listed twice. Again the protoforms are Wang^s (1979) PHM, and the resulting forms were given for the Yanghao dialect, but given the relative uniformity of the [s] and [sh] series in Table 12a, they are in this thesis analyzed as the PQH phonemes as well.

2 What Wang (1979) here notes as the voiceless alveolo-palatal fricative [G] is listed in Ma & Tai (1956) as a palatal fricative [9] and will be treated as such in the remainder of the reconstruction. The alveolo-palatal *nt§(h) and *§(h) items are also interpreted as palatal in this reconstruction. Of course, further fieldwork would help clarify this phonetic detail. 74

(22) PHM *nts- > s-

PHM *ntsh~, -tsh > sh-

PHM *ntg- > G-

PHM *nt§h- > G'1-

(23) PHM *s- > sh-

PHM *§- > GH-

PHM *nts- > s-

PHM *nt§- > G-

As mentioned, both of these sets of changes have analogues attested elsewhere.

The analogue to (22) is the development of aspirated fricatives in Burmese. Between

Middle (1400s) and Modern Burmese, a chain shift occurred beginning when Old/Middle

Burmese s- became a dental affricate (Jacques 2011:8). Once this happened, the alveolar affricates lenited to fricatives, and maintained their aspiration contrast. Relevant changes are listed below in (24). 75

(24) s > [t0]

c [ts] > [s]

ch [tsh] > [sh]

kr-, ky- > [tc]

khr-, khy- > [teh]

(Jacques 2011:7)

There are also intermittent, lexically determined cases of a change analogous to the above known in languages that already have a separate origin for existing aspirated fricatives. One example is that the Amdo Tibetan reflex of the Old Tibetan adverb [tshur]

'towards oneself is [s''3r], not the expected [tsh3r] (Haller 2004:39). However, this comes as an addition to a regular change producing aspirated fricatives as part of a process of cluster simplification.

As far as the changes in (23) are concerned, a similar change is known from the southern dialect of Shan (Edmondson 2008:197). The changes listed below in (25) occurred in southern Shan.

(25) *s/z > sfl

*te/dz > s-

Interestingly, northern Shan seems to have kept the contrast between the two, but 76 instead has [ts] and [s] in place of [s] and [sfi]. Northern Shan, in other words, retains a fricative versus affricate contrast, whereas the southern variety has transformed it into an aspiration distinction. This is demonstrated with relevant forms in Table 13 on the next page. Edmondson (2008) does not discuss any information that would indicate which part of the shift came first, but given that the Burmese chain shift was a pull shift, starting with the fricatives, the assumption in this thesis is that the Shan shift was as well. The result of this assumption is that both types of change in Qiandong can be posited as starting with fricatives, rather than the more complex alternative of positing different motivations for the two change types in Qiandong.

Table 13: Shan Data

[Sibilant source Gloss Northern Shan Southern Shan *cfe boy, male tsaai2 saai2

! *ck elephant tsaaq4 saaq4 *ck artisan tsaaq6 saar)6 *tch prince, lord tsau3 sau3 *s three saaml sfiaaml i *s four si5 sfii5 *s unmarried girl saul sfiaul i *z wash (hands) sak8 sfiak8 ; *z wash (surface) suk8 sfiuk8 • *s tiger suil sfiuil

fi ; *S pillar saul s aul *s tall, high sugl sfiugl (Edmondson 2008:196, Li 1977) 77

The initiation of the Qiandong chain shift is still uncertain. The Burmese shift was a drag chain; as such, the analogy between the Burmese and Qiandong shifts is tighter if the Qiandong chain shift is also a pull chain (Jacques 2011:8). However, the aspiration of the [s] is a strange change taken on its own, whereas it can be motivated from the top as the unaspirated prenasalized initials simplify if the Qiandong chain shift is a push chain. The aspirated cases would then have followed suit to preserve the symmetry of the system, what Hock (1991) terms a solidarity chain. Without more evidence, it is not clear which type of chain is the superior analysis, but neither has significant repercussions for the rest of the thesis, so the question can remain unanswered for the time being.

An added benefit of adopting Wang (1979)'s analysis here is that positing prenasalizations dramatically simplifies the picture in Qiandong with regard to palatal consonants, among others. For instance, the majority unaspirated and aspirated palatal stop series, series 3 and 8 in Table 4a at the beginning of Chapter 5, have a few affricate reflexes, as seen in Table 7a, but there are also majority palatal affricate series, series 4 and 6 in this chapter's Table 12a, with stops as reflexes in Sansui, shown in Table 7b.

Tables 4a, 7a, and 7b are repeated on the next two pages for the reader's convenience. 78

Table 4a: Plosives in Ma & Tai

Dialect Series Series Series Series Series Series Series Series Series Series 1 2 3 4 5 6 7 8 9 10

Congjiang P t c k q Ph th ch kh qh Huangll p/p" t/th c/ch k/kh q ph th ch kh qh Jiaba P t c k q p'1 th ch kh qh Jinping P t *9 k q P" th t?h kh qh Jingxian P t t? k q ph th t£h kh q" JTnzhong P t c k q Ph th ch kh qh Jiuzhou P t c k q ph th ch kh qh ... Paiting P .t c k q p'1 th ch kh qh Sandu P 1 c k k Ph th ch kh kh Sansui P t c k q Ph th ch kh qh Taigongzhai p t c k q Ph th ch kh qh Taiyong P t c k q Ph th ch kh qh Wucha P t c k q Ph th ch kh qh Wuluo P t c k q Ph th ch kh qh XTnqiao P c k q ph th ch kh qh Xuanwei P t c k q ph th ch kh qh Yong'an p/ph t/th c/ch k/kh q ph th ch kh qh Zhenyuan P t c k q ph th ch kh qb ZhouxT P/Ph t/th c/ch k/kh q PH th ch kh qh (Ma & Tai 1956:43-45) 79

Table 7a: Prenasalized Stop Series

Gloss Jingxian Taigongzhai PQH

h 'change' [t9hu3] [che3] *nc 'chopsticks' *nc [t?o6] [cou^] 'crab' *nc [t9au'] [CO1]

h 'shave' [t9"i5] 1 [chi5] *nc

Table 7b: Non-prenasalized Stop Series

Gloss Sansui Taigongzhai PQH

h cloth [chei]'] [t^u1] *c

'medicine' [ca1] [t^a1] *c

h 'money' [che^] [t9he^] *c

'succeed' 2 2 *c tee ] [t9arj ]

h 'wipe' h h 2 *c [c e^] [t9 e ] (Ma & Tai 1956:37)

If reconstruction simply abides by majority rule here, a typologically bizarre and seemingly inexplicable change of fricative to stop needs to be posited in Sansui. Instead, if we posit the palatal stop series in Table 7a as formerly prenasalized stops, the palatal affricate series in Table 7b as non-prenasalized stops, and the palatal affricate series as affricates which may or may not be prenasalized in Table 14 on the next page, and assume reduction in all cases, the contradiction is resolved, as both 7a and 7b can give rise via reduction to affricate and stop forms in different dialects, and the majority- 80 affricate series reduce to a fricative in the appropriate dialects.

Table 14: Series 12 and 13, Reconstructed as Affricates

Gloss Gaotongzhai Taigongzhai Taiyong PQH

'end' *(n)tsh

erase *(n)t9h

'song' [?i7] *(n)t9h 'thousand' [shaq] [tshaq] *(n)tsh (Ma & Tai 1956: 36)

The data in Qiandong seems to support this analysis. In particular, the Burmese- style change pattern in (22) would expect prenasalized affricates to denasalize, but only some of those to continue leniting to become fricatives, and that seems to be what's happened. Moreover, Yanghao words beginning with [s] and [sh] typically seem to have cognates in other Hmongic languages outside of Qiandong that begin with [nts] or [ts] for the former, and [ntsh] or [tsh] for the latter (Ratliff 2010 :71-73).

The change of the Shan type in (23) is overlapping in Qiandong with the change of the Burmese type in the reduction of prenasalized affricates to fricatives. The only thing that still needs evidence is the development of the fricatives into their aspirated counterpart. Indeed, Purnell (1970:46-48) details two distinct series with [sh] reflexes in

Qiandong; one such series has primarily affricate reflexes elsewhere in Hmongic, which would be tied to the Burmese-type change, while the other has mostly [s] reflexes, and 81 would be associated with the Shan-type change.

Wang (1979), as can be seen in (21), has PHM *tsh reducing to a fricative as well, but this may be deceptive for the purpose of PQH reconstruction. Series 3 in Table 12a, dominated by alveolar affricates, also has fricative and palatal affricate reflexes. Neither would make it particularly problematic to reconstruct alveolar affricates for this series in

PQH. However, if PHM alveolar affricates have reduced or are reducing by PQH, then there is no clear way for a PQH alveolar affricate to arise. However, Wang (1979)'s work was only meant to apply to Yanghao; if Yanghao has a fricative reflex for this series, then this is not a contradiction. The PQH alveolar affricate is indeed the PHM alveolar affricate, but it happens to have reduced to a fricative in the dialect Wang (1979) worked on. This analysis is supported by the fact that, as stated on the previous page, Yanghao's

[s] and [sh] series have affricate reflexes elsewhere in Hmongic. As such, here *ts and

*tsh are reconstructed for PQH, shown in Table 15 below.

Table 15: Alveolar Affricates

Gloss Jiuzhou Taigongzhai Wucha PQH

'fish basket' [t9ha^] [tsha^] [sha^] *tsh 'five' *ts [tfa1] [tsa1] [sa1]

'house' 3 3 3 *ts [tPi ] [tse ] [sa ]

'stamp' h 5 h h h [t? o ] [ts o^] : [s D^] : *ts 82

7.2 Labiodental Fricatives

The [f1] and [f] series remain remarkably distinct in terms of aspiration, and thus can be reconstructed as-is for PQH. However, this brings up the question of how they might have developed. In this section, it is argued that a different process from those in subsection 7.1 is responsible; rather, the series developed from a pre-Qiandong *pi and

*pjh_

First, applying the previously described changes of the preceding subsection seems implausible. One could, conceivably, reconstruct an *npf'/*npf or *pfV*pf that reduces to a fricative; some such reduction would be necessary in either of the pathways posited so far. However, this faces at least two problems. One is that unlike the previously mentioned prenasalized affricates, such a sound would have no counterparts anywhere else in Hmongic. What's more, along the Burmese-analogue path of change, one would expect the prenasalized aspirate to leave some aspirated affricates as reflexes yet there are absolutely no labiodental affricates to be found.

This could be circumvented by assuming only an unaspirated *npf or *pf and a

Shan-type change, but this then begs the question, why would only the labiodental place of articulation lack an aspiration contrast in affricates? While perhaps an excusable issue, as natural languages can and do have such gaps, the cumulative effect of this reasoning discourages pursuing the origin of the [f1] from the standpoint of previously posited types of change. One might instead simply posit an *f that gained an aspiration distinction in analogy with the other places of articulation. Perhaps after aspiration distinctions developed in other fricatives, *f split sympathetically in some environment; this would be another solidarity chain shift. Conventionally one would seek to reconstruct an environment for the split, but under the exemplar view such an environment might be as simple and yet opaque as a difference in word frequency of use. However, it would require the abandonment of our methodological preference for reconstructing

Neogrammarian-type change, and thus is a distinctly suboptimal solution.

In seeking an alternative, we turn to the literature, specifically to other languages and ways they have developed an aspiration distinction in labiodental fricatives. Jacques

(2011:4) lists three languages and one reconstructed proto-language with [f] and [f1]. Out of these, Heqing Bai seems in published work not to have a clear path of evolution of that distinction, while one of the remaining languages is this work's own Yanghao; neither will be helpful here.

This leaves the Siouan language Ofo (Rankin 1988) and Pulleyblank's (1984a) reconstructed Late Middle Chinese. Ofo's aspirated fricatives were gained by the fusion of a laryngeal element to fricatives (Jacques 2011:12). Such a development would be very strange in Hmongic; as Hmongic lacks coda consonants other than [rj], the environment would have to be present in existing onset clusters at an earlier stage, and it would beg the question, why is there no evidence of this laryngeal fusion anywhere else in Hmongic? Ofo's pathway does not seem viable for our purposes. 84

The last option provided by the literature, the case of Late Middle Chinese

(LMC), however, is much more helpful. It seems that in LMC *f and *1* developed from

Early Middle Chinese *p> and *pjh, respectively, retaining the aspiration distinction. The probable phonetic explanation for such a change is that the movement of the tongue towards the palate was retimed to be simultaneous with the stop, breaking up the stop closure. Such a pathway is highly appealing to explain [f] and [f1] in modern Qiandongic dialects. Generally, diachronic change from a bilabial stop to a labiodental fricative is quite common, as in Grimm's Law, and a notable synchronic analogue for the process at hand is, of course, the well-known Irish initial lenition process' inclusion of a spirantization of a palatalized bilabial.

Such a change also fits well with what is known about Hmong-Mien broadly; palatals and palatalized onsets are common in the family. Indeed, a palatalization distinction continues to exist in laterals, as descripted in Chapter 8, so adding another is quite natural.

Finally, the palatalization allows a bilabial source to be posited for both the [p] and [ph] series in modern Qiandong and the [f] and [f1] series without violating the principle of Non-Duplication in the methodology. The distinction between palatalized and nonpalatalized bilabials was transformed into one of manner of articulation as the palatalized bilabials lenited, without loss of the contrast.

Some caution is warranted in drawing on proto-languages as evidence of a particular sound change being valid, as they are subject to revision in a way that attested 85 language data is not. The evidence for contrastive *f and *fh in LMC is primarily philological; Chinese rhyme tables indicate the two existed (Jacques 2011:4). On purely reconstructive evidence, a *pf/*pf' reconstruction is also plausible for LMC, but one would then be forced to explain the rhyme table evidence in some other manner, and again, a development of *p> and *p'h to *f and is a natural one. Moreover, this sound change is similar to that of Burmese and a wide variety of other attested languages in that it involves an aspiration distinction existing in something other than fricatives in the proto-language, and that aspiration distinction being maintained as the original phonemes become fricatives (2011:20). Thus, it can be seen as fitting a larger pattern in aspirated fricative development.

Drawing on this, if we posit a *pi and *p'h in a pre-PQH stage, which evolve to *f and *fb in PQH, then there are no typological difficulties in positing *f and *fh in PQH.

The next subsection explains the non-conforming reflexes in those series.

7.3 Aspiration Mergers

As with stops in subsection 5.2, mergers of fricatives to an aspirated state are restricted by tone. Aspiration occurs for all voiceless unaspirated fricatives, including [f],

[s], [9], and the lateral fricatives [1] and [lj], which are discussed in the next chapter, in tone D2 words in ZhouxT; it also occurs for [f] and [s] in tone A2 in Wucha, Zhenyuan and ShTdongkou. Table 16 displays the aspiration merger concerning tone D2 in ZhouxT, 86 while Table 17 contains examples of the aspiration merger in tone A2 in the other dialects. Both tables are shown below.

Table 16: Tone D2 Fricative Aspiration Merger

Gloss Taigongzhai Zhouxi PQH

'drag' 8 7/8 [l'o ] [}jh0 ]

'harelip' [fi8] [Pi7/8] *f

'a piece' l«a8] |l"e7/8] *1

'wet' [?e8] [9hau7/'8]

Table 17: ToneA2 Fricative Aspiration Merger

Gloss ShTdongkou Taigongzhai PQH

'bright' [Pag2] [far)2] *f

'hand up' [Pi2] [fi2]

'meeting' [shou2] [sou2] *s

'money' [shei2] [sei2] *s (Ma & Tai 1956: 38-39)

In the ZhouxT case, this is the same tone in which unaspirated stops aspirate, as discussed in subsection 5.2, and a shared origin for both aspirations is a near certainty.

In the tone A2 case, a separate origin is needed, but tonal phonation is still the presumed cause. 87

The dialects in which both the PQH *f and *f' series have unaspirated reflexes

can also be treated as mergers. This is shown below in Table 18 below.

Table 18: Aspiration Loss in PQH *1*

Gloss Congjiang | Taigongzhai PQH

'belight' [fe1] Ve1] *f'

'rub in hands' ^lj IfPa1] ' *fh

(Ma & Tai 1956: 35)

As aspirated fricatives are unstable sounds in general, the loss of such a contrast is

far from unexpected (Jacques 2011:2-3, Wheatley 2003:199).

7.4 Fricative-Aspirated Velar Stop Alternations

In two fricative series, the predominantly [x] and [

aspirated velar stop reflexes. The [911] case is the more complex, so first the [x] case is

considered. Examples are given in Table 19 below.

Table 19: Velar Fricatives and Aspirated Stops ; Gloss Shidongkou Taigongzhai PQH

'new' [kM1] [xi1] *x

| 'quick' [ khj5J [xa5] *x (Ma & Tai 1956: 37) 88

Positing a change from fricative to aspirated stop here would be typologically bizarre and difficult to motivate, but there is already an exceptionless correspondence set for the aspirated velar stop in the data in Ma & Tai (1956:41), and no environment for a split is apparent, so a lenitive change in all other reflexes is not an appealing solution either. An appeal to prenasalization and lenition would be tempting if not for the fact that in Purnell (1970:49), Qiandongic velar fricatives correspond to alveolar fricatives elsewhere in Hmongic, not anything like a prenasalized cluster.

Something to note is that unlike other fricatives, no aspiration distinction is apparent in Qiandongic velars; Purnell (1970) records an aspirated velar fricative in a few languages' inventories, but never contrasting with an unaspirated counterpart, and as such it is unclear what his recordings were intended to convey. This gap is not likely a coincidental peculiarity of Qiandongic languages. In the Shuiluo dialect, spoken in

Sichuan's Muli County, of the Tibeto-Burman language Pumi, there is a pattern of diachronic lenition processes that, had it been followed in velars, would have produced the following pair of changes in (26).

(26) *skh- > xh-

*sk- > x-

Instead, in modern Shuiluo Pumi, both have merged to x- (Jacques 2011:29). 89

Also, in Chumashan, the obstruent dissimilation effect that typically produces aspirated fricatives at morpheme boundaries does not apply to velar fricatives (Klar 1977:11-14).

Jacques (2011:30) suggests that the aspiration contrast in velar fricatives is difficult to maintain for phonetic reasons, writing, "aspiration and friction are less distinct for velars than at other places of articulation; velar fricatives have more energy in the lower range of the spectrum than other fricatives, and resemble aspiration."

This suggests there may be a perceptual rationale for positing a change from *x to kh. Given the existence of a separate *kh series, perhaps the fricative manner was mistaken as aspiration on a stop, and the two series merged. This is the interpretation held to in this thesis, and *x will be reconstructed for PQH despite the stop reflex.

Something similar is happening with aspirated palatal fricatives in Jiaba; the Jiaba reflexes are either unaspirated palatal fricatives or aspirated velar stops. This is shown in

Table 20 below. 90

Table 20: Aspirated Palatal Fricative Alternations

Gloss Jiaba Taigongzhai PQH

h 'blood' [9haq3] *9 [-back]

h 'fear' [khi!] [?ha ] g [+back]

h 'song' [kha7] [?ha7] *g [+back]

sour [90 M [q'-O1] [-back]

(Ma&Tai 1956: 40)

While the explanation for velar fricatives was dependent upon their phonetic qualities, the same analysis can in fact be extended to the palatals in Jiaba. If there was an allophonic difference in backing in PQH palatal fricatives caused by the backness of adjacent vowels, analogous to that in German, the back allophones would be velar, and thus susceptible to the same kind of perceptual confusion that produced stops in the velar fricative case. The forward allophones would remain palatal, and the deaspiration can be analyzed as a mere collapse of the aspiration distinction; as mentioned in subsection 7.3, the aspiration distinction in fricatives is highly unstable and this is thus extremely plausible. The modern rimes do not show a backness distinction, but the modern rimes in the Qianddng data as it stands show few discernible patterns at all, suggestive of significant change in the rime system that could conceal an earlier vocalic distinction in backness. Furthermore, effects of absent vowels are also part of the analysis of the next 91 subsection, so assuming vowels that are no longer present aid the reconstruction in more

than one way. Thus, the reconstruction of *9h for PQH is maintained here, but allophony

in backness is reconstructed.

7.5 Voiced Velar Fricative Fronting

Series 14 in Table 12a has a plurality of [y] reflexes, but also has [j] as a reflex, as

well as combinations of [j] and [v] and of [y] and [w]. Relevant examples are given below in Table 21.

Table 21: Series 14

Gloss Jingxian Taigongzhai Zhenyuan 1 Zhduxi PQH I N — < _ - T L o C 'dragon' [voi]2] [yog2] - > *Y

'guard' [vou3] [ye3] - [wo3] *Y

'stone, rock' 1 1 *y Dei1] [ye ] Vi [yi ] 'vegetable' [jau1] [yo1] [jo1] ! [wau 1 ] *y (Ma&Tai 1956: 39)

As PQH *j has already been reconstructed for the exceptionless [j] series, it can safely be ruled out as per the Non-Duplication principle. Majority Rule suggests [y] as the proto-form, but a shift from [y] to [v] would be highly aberrant.

It is notable, however, that all of these are labial, velar or palatal. One possibility is that Majority Rule can be more or less maintained if the proto-form is in fact *y, but accompanied by a labialization of the consonant or roundedness on a succeeding vowel or onglide. As a vocalic no longer present was posited in subsection 7.4 previously, reconstructing another here does not add unnecessary complexity to the reconstruction.

This is a very interesting possibility, as evidence of a disappearing rounded onglide was precisely the characteristic Ratliff (2010:24) found as a commonality between Pa-Hng and the XiangxT language Qo Xiong. If this is indeed evidence of the same rounded onglide, and that is by no means a foregone conclusion, then it suggests that this onglide was lost later than the split of Qiandong and XiangxT. A search for evidence of this onglide in Chuanqiandian and Jiongnai/Ho-Ne might then determine when it was lost in each branch and what the ramifications are for familial substructure.

Finally, if there is in fact a labialization or rounded vowel or onglide at work here, the palatal [j] can then be explained as the result of a coarticulation of the rounded or labial element and [y].

In cases where *y was not followed by such an element, the reflexes might have disappeared from view in any number of ways. For instance, without the labial element providing a forward component, they might have fallen back to [h], either merging with the existing laryngeal series or, if the change early enough in PQH development, in fact created it, becoming the PQH *h series. This is by no means intended as a definitive explanation, but shows that the lack of obvious reflexes for *y without labialization in modern dialect need not be a cause for concern. However, it is notable that as this series 93 consistently has a palatal reflex in Zhenyuan, it can not be the case that not all members of Series 14 had the labializing component if the explanation in the preceding paragraph of the presence of [j] is to hold.

It is still unclear why the reflexes would be inconsistent, as shown in Table 21, but that may also be a result of vocalic developments. Reconstructive work on the PQH vowel system is sorely needed to determine whether this is indeed the right analysis for

Series 14, but on the basis of the arguments presented here, *y is chosen as the reconstructed phoneme for Series 14, and is presumed to be accompanied by a rounded element.

7.6 Summary of Fricative and Affricate Developments

Figure 6 below details reconstructed PQH onsets posited in this chapter. In order to justify reconstructing an aspiration distinction in labiodental fricatives, *pJ and *p>h were also posited in 7.2 for a pre-Qiandong stage of development. (27) lists changes posited.

Figure 6: Summary of PQH Onsets Reconstructed in Chapter 7

*s *9 *j *x *y *h

i *ts *tsh *t? *t9h *nts *ntsh *nt9 *nt9h ! 94

(27) PQH {*v, *h, *j} no change

Qiandong Chain Shift:

PQH *nts(h)- > ts(h)-, s(h)- in differing dialects

PQH *nt9(h)- > tfC1)-, 9(h)- in differing dialects

PQH *s- > sh-

PQH *9- > 91'-

Palatalized Bilabial Spirantization - Pre-Qiandong *pi(h) > PQH *f(h)

Zhouxi Tone D2 Spirant Aspiration - PQH [+continuant, -sonorant, - voice] >

[+spread glottis] in tone D2 reflexes in Zhouxi

Tone A2 Spirant Aspiration - PQH {*f, *s} > [+spread glottis] in tone A2 reflexes in Wucha, Zhenyuan and Shldongkou

Labiodental Fricative Deaspiration - PQH > f in Congjiang, Jlnpi'ng, Jingxian,

Paiting, Sandu, Sansui and Taiy5ng

Velar Fricative Strengthening - PQH *x > kh in Sandu, Zhenyuan and Shldongkou

Allophonic Palatal Fricative Deaspiration and Velar Stop Formation - PQH *9h >

9 and kh in Jiaba

Voiced Velar Fricative Fronting - PQH *y[LABLAL] > v and j intermittently in

Zhouxi, w sporadically in Zhouxi, and j consistently in Zhenyuan, no change in all other dialects 95

8.0 Laterals

There are five lateral series in the Ma & Tai (1956) data, and eight in Purnell

(1970). Given the large number of lateral series, the development of laterals has been given its own chapter. All are detailed in Tables 22a and 22b below.

Table 22a: Laterals in Ma & Tai

Dialect Series 1 Series 2 ! Series 3 Series 4 Series 5 Congjiang 1 L> i/i 1/1 Huangli 1 V \h lJh 1 Jiaba i 1L' | ih JTnping 1 l1 i/i Jingxian 1 P I V JTnzhong 1 1' ih }jh j Jiuzhou 1 IP h jjh i » Paiting 1 V i/i 1 Sandu 1 : V I V 1/1 Sansui 1 1j I V 1 Taigongzhai 1 1j ih 1 ! Taiyong 1 1' i/i lih 1 I Wucha 1 1' ih }jh i i Wuluo 1 P h }JH i i XTnqiao 1 \- ih }JH 1 i Xuanwei 1 1' I lJ 1 | Yong'an ;1 \h l'h Zhenyuan 1 1' \h Jjh i :

h h Zhouxi 1 lJ i Ijh i/i 96

(Ma & Tai 1956:43-45)

Table 22b: Laterals in Purneil

Dialects Ser. 1 Ser. 2 Ser. 3 Ser. 4 Ser. 5 Ser. 6 Ser. 7 Ser. 8

Gaotongzhai 1 1 1 - - \ 1

Kaitang 1 1 - - 1

Shldongkou 1 1 lh 1 1 1

h Yanghao - - il - - 1 1 | Zhenfeng 1 1 Th lh lh lh lh 'lh (Purnell 1970: 50, 54-56)

In Table 22a, each series seems to be dominated by a particular reflex: [1], [P], [lh],

[l|h] and [1] respectively, with Series 1 and 2 being in fact exceptionless. Figure 7 below clarifies data from Table 22a to show that these five forms are in fact the only extant reflexes of laterals in Qiandong.

Figure 7: Lateral Reflex Attestation in Qiandong

; i Unaspirated Aspirated Unpalatalized Palatalized Unpalatalized Palatalized

I Fricative Attested Unattested Attested Attested Liquid Attested Attested Unattested Unattested

This is taken as evidence that all five series can be reconstructed with those forms 97 at PQH level, providing *1, *]->, *1, *lh, and *JJh. Subsection 8.1 discusses deviations from those forms in particular modern dialects, subsection 8.2 discusses the development of aspiration contrasts in lateral fricatives, and subsection 8.3 summarizes the results of the previous two.

8.1 Reflexes Changed From PQH

The first notable difference from the posited PQH forms for each series is the unaspirated liquid reflexes in four dialects of the two *l(->)h series. This is shown below in

Table 23.

Table 23: Lateral Deaspiration and Liquefaction Merger

Gloss Jingxian Sandu Sansui Taigongzhai Xuanwei PQH | *ljh 'big' i[lJou!] [lJau' ] [ki1] [lih1]3 [in1] j *Jh 'iron' [lou^] [lou5] [ho5] [lhau5] [ID5] *Jh 'moon' [lei5] [la5] [lue^] [lha5] [la5] (Ma & Tai 1956: 36)

In Jingxian, Sandu, Sansui, and Xuanwei, the distinction between PQH *1 and *1J on the one hand and *lh and *ljh on the other appears to have collapsed in favor of the unaspirated liquid. It may be that the fricative and the unaspirated liquid were close

3 Final missing in original (not translated) Ma & Tai text 98 enough acoustically for the perceptual distinction to become muddled and collapse.

Moreover, there may, as in subsections 5.3 and 6.4, be a distantiation rationale here; five laterals is a fairly large set, and crowding in the perceptual space may play into the merger by reducing that number to three.

Another aberrant reflex is Jingxian [s] as a reflex of PQH *1. Examples are given below in Table 24.

Table 24: Jingxian Delateralization Merger

Gloss Jingxian Taigongzhai PQH

'beg' [sa^] [le5] *1

'four' [sau1 ] [lo1] 1 1 I*1 (1956: 36)

Given that [I] and [s] are very similar articulatorily, a merger of the two is not very strange, [s] is the more preferred sound, and is more cross-linguistically common, so the merger operating in that direction is also not unnatural. Again, distantiation may be a factor, particularly since Jingxian is one of the dialects affected by the first merger in this subsection. Assuming that both mergers began when there were still five laterals, distantiation may have acted on all three affected phonemes to relieve the perceptual pressure.

ZhouxT's aspiration distinction in the PQH *1 is explained as part of a tonally contextual aspiration effect in subsection 7.3. This leaves only the appearances of [1] and 99

[1] as alternating reflexes for PQH *lh in Congjiang, Jinping, Paiting, and Taiyong, and for *1 in Congjiang and Sandu. These are shown in Tables 25a and 25b below.

Table 25a: Alternating *lh Reflexes

Gloss Congjiang Jinping Paiting Taigongzhai Taiyong PQH

'marrow' [Ir1] [lue1] [la1] [f'e1] [le1] : *Jh 'moon' [le5] [H5] [la5] ; [lha5] , [le5] *Jh 'string, [la5] [la5] [la5] [ihe5] [le5] rope'

Table 25b: Alternating *1 Reflexes

Gloss Congjiang Sandu Taigongzhai PQH

'beg' [le5] [le5] [le5] *1 j 'four' [lau1] [la1] [lo1] *1 (1956: 40)

Ma & Tai (1956) did not find any predictable synchronic distribution or conditioning factors to these reflexes. However, given the small number of lexical items in Ma & Tai (1956) and the large number of dialects researched, this may be simply a result of the paucity of data available, and be explicable either with a better understanding of QiandSng rime development, or with more detailed fieldwork on the contemporary dialects. This thesis proceeds with the assumption that these lexical items 100 will have been found to have undergone an explainable change from the lateral fricative proto-forms already posited, but if future research supports the lack of predictability found by Ma & Tai (1956), a reexamination may be in order.

8.2 Lateral Fricatives

As with other fricatives previously discussed, positing aspiration distinctions in lateral fricatives warrants reflection on how such distinctions could have come about.

None of the previously discussed pathways that lead to aspiration contrasts in spirants are adequate to explain developments in laterals.

The Shan or Burmese types of change discussed in subsection 7.1 on the surface seem like a viable solution. Either scenario requires the postulation of some kind of cluster involving the lateral. Such clusters are widespread elsewhere in Hmongic

(Purnell 1970:54-56), so postulating them in the Qiandongic branch is not bizarre. More conclusively, many of the series in Purnell (1970:50,54-56) have an obstruent-lateral cluster as a reflex at least some of the time in cognates in other Hmongic branches, which is demonstrated in Table 26 on the next two pages, an extended version of Table 22b. 101

Table 26: Lateral Correspondences With Other Hmongic Branches

Dialects Purnell name Subgroup Series 1 Series 2 Series 3 Series 4

h YSnghao MKL = Miao Qianddng l - of K'ai-li Zhenfeng MCF = Miao Qiandong 1 1 lh lh of Cheng-feng Shldongkou MTK = Miao Qianddng 1 1 ilh 1 of Tai-kung

Kaitang ! MLS = Miao Qiandong 1 1 i - - of Lu-shan j

Gaotongzhai MJC = Miao Qiandong 1 1 ! 1 - of Jung-chiang i j Layiping MHY = Miao XiangxT : ?1/1 i 1 lh of Hua-yuan

h Shuiwei MLL = Miao Chuanqiandian ?1 ; 1 l - of Lung-li

h Shimenkan MWN = Miao Chuanqiandian ?1/1 l - of Wei-Ning Thailand MPT = Miao Chuanqiandian 1 lh pl White of Petchabun Hmong Dialects Purnell name Subgroup Series 5 1 Series 6 Series 7 Series 8

Yanghao MKL = Miao Qiandong 1 1 - of K'ai-li Zhenfeng MCF = Miao Qiandong lh lh lh lh of Cheng-feng

Shldongkou MTK = Miao j Qiandong - 1 1 - of Tai-kung

Kaitang MLS = Miao Qiandong ! - 1 - - of Lu-shan

Gaotongzhai MJC = Miao Qiandong 1 |1 - of Jung-chiang

Layiping MHY = Miao Xiangxi qw ; pr -/br j of Hua-yuan | 102

Shuiwei MLL = Miao Chuanqiandian ql pi of Lung-li Shimenkan MWN = Miao Chuanqiandian thl tl tl of Wei-Ning Thailand MPT = Miao Chuanqiandian phi ?d pi pi White ofPetchabun Hmong

The problem lies in the fact that any chain shift like those involving alveolars and palatals in Burmese, Shan and Qiandong needs to be actuated from the bottom.

Preexisting simple laterals would have to lenite to zero, aspirate, or be removed from the system through some other change to start the chain shift and cause reduction of obstruent-lateral clusters. This is problematic because there is no evidence in the Purnell

(1970) data that any series with predominantly lateral reflexes elsewhere in Hmongic has deleted or become something nonlateral in Qiandong; in other words, there is no actuation mechanism for the pull chain shift, as the laterals never leave the system the way [s] and [9] do for the alveolar and palatal chain shift. Moreover, the series that have obstruent-lateral clusters in other Hmongic branches (1970:54-55), while often leniting to plain laterals, do not clearly retain their aspiration distinctions, a critical feature of a

Burmese style development. On the other hand, Shan type change would mean that an unaspirated fricative had become aspirated, and one would expect to see unaspirated reflexes outside of Qiandong to an [lh] series within it. That is not, in fact, attested in

Purnell (1970). As such, the Shan pathway also is lacking in evidence. 103

Labiodental fricatives were proposed in subsection 7.2 to have originated from

palatalized bilabial stops and retained an aspiration distinction from that state. There is

no clear way to apply such a pathway to laterals.

Thus, a new pathway must be posited. Figure 7 below, repeated from 8.0, shows

the logical possibilities, given the three parameters of palatalization, manner, and articulation. Notably, the aspirated liquid variants are effectively unattested.4

Figure 7: Lateral Reflex Attestation in Qiandong

Unaspirated Aspirated Unpalatalized Palatalized Unpalatalized Palatalized Fricative Attested Unattested Attested Attested Liquid Attested Attested Unattested Unattested

Knowing this, we can turn to another type of change in Jacques (2011:9), spirantization of aspirated sonorants. It is not uncommon for aspirated sonorants to spirantize, particularly [jh], [rh] and [wh] (recall from subsection 4.2 the equivalence of voicelessness and aspiration in sonorants), and there are two known cases of them becoming aspirated.

One such case is Cone Tibetan, where Cone Tibetan [§he] has been traced to Old

4 To be more precise, Zhenfeng is listed in Purnell (1970:18) as having an aspirated liquid, but this is suspect. Purnell (1970)'s Zhenfeng data came from Esquirol (1931), a dictionary criticized by both Kwan (1966:17-18) and Purnell (1970:17) as imprecise and opaque. Purnell in particular writes that "Esquirol's descriptions lack the precision needed to elucidate the problem areas" (1970:17). Thus, it is quite possible that the fricative versus liquid distinction was simply missed by Esquirol (1931). This is supported by the fact that lateral fricatives appear nowhere in his work. 104

Tibetan hral, the initial of which was probably *rh. /§h/ is a rare initial in Cone Tibetan and elsewhere in the language its origin is unknown. The second case of note is in Proto-

Mazatec, where Jacques (2011:10) reinterprets Kirk (1966)'s *P as *jh becoming [j11] in modern Chiquihuitlan Mazatec based on data from more recent works (Jamieson 1966,

Rensch 1976, Silverman et al. 1995).

Given that aspirated liquid laterals seem to be absent from the data, the possibility rears its head that something similar has occurred here, namely that *lh at some pre-

Qiandong stage developed into PQH *lh. This is consistent with Figure 7 and has the added benefit of being consistent with Series 3 in Table 26. Moreover, in both known cases of voiceless sonorant spirantization preserving aspiration, the languages involved already had an unaspirated version of the relevant fricative, and other aspirated fricatives created by other means. This is supportive of this analysis, as PQH has the other fricatives created by the phenomena discussed in previous chapters.

Finally, this allows for an explanation of a strange gap in Qiandong and posited

PQH laterals. PQH *ljh is reconstructed and [P1] is present in the modern dialects, but [I1] is found nowhere and not reconstructed. Yet, as the unaspirated fricative, it should be the more preferred form.

The other part of this quandary is tied to the origin of the palatalization distinction. Often, palatalization comes from an adjacent high vowel or glide, as in Irish.

If such a vowel or glide was at work in this instance, and later deleted, a split could be 105 posited in pre-Qianddng\ If this split occurred before the spirantization of the aspirated liquids, and only affected liquids, the unaspirated fricative would never have been affected by the palatalization, but the future aspirated fricatives would have been. The aspiration on the palatal lateral fricative, the one with no unaspirated counterpart, was presumably retained because fricative aspiration is so prominent in Qiandong otherwise.

On the following page, figures 8a and 8b show first and second relevant prior stages of development under this analysis, while Figure 7 is repeated again as an end result. That is to say, the three figures are chronologically ordered.

5 Ma & Tai (1956:47) give a handful of words in which a nonpalatalized lateral coexists with a high 2 vowel, such as 'moon' [lye ] in Xuanwei, but this can be explained by the existence of a non-high element, probably a vowel or an offglide, intervening, and deleting at the same time as the high elements, as the palatalization was phonologized on the consonant. This missing vowel can be posited with relative confidence given that vowels were similarly used in arguments in subsections 7.4 and 7.5 of the analysis. Alternately, the high vowels may have been created after the palatalization by subsequent vocalic changes. In either case, further work on PQH rimes would be revealing. 106

Figure 8a: Pre-Palatalization State

Unaspirated Aspirated Fricative Present Absent Liquid Present Present

Figure 8b : Liquid Palatalization

Unaspirated Aspirated Unpalatalized Palatalized Unpalatalized Palatalized Fricative Present Absent Absent Absent Liquid Present Present Present Present

Figure 7: Lateral Reflex Attestation in Qiandong

Unaspirated Aspirated Unpalatalized Palatalized Unpalatalized Palatalized i Fricative Attested Unattested Attested Attested Liquid Attested Attested Unattested Unattested

As can be seen in Figure 8a, the only proto-forms needed for the initial pre-

Qiandong state are then *1, *lh and *1. This explanatory power is taken as justification for reconstructing an aspirated liquid spirantization pathway leading into PQH. 107

8.3 Summary of Lateral Developments

Figure 9 below details reconstructed PQH onsets posited in this chapter.

Furthermore, refer to Figures 8a, 8b and 7 in subsection 8.2 for insight into two pre-

Qiandong developmental stages. (27) on the next page lists changes posited; the unclear changes in Tables 25a and 25b are omitted, as the pathway of change is unclear.

Figure 9: Summary of PQH Onsets Reconstructed in Chapter 8

*1 : *jj *1 *lh

(28) Early Pre-Qiand5ng *l(h)[+high] > Late Pre-Qiandong *l'(h)

Late Pre-Qiandong *l(->)h > PQH *l(J)h

PQH {*1, *P} no change

PQH *l(J)h > 1(J) in Congjiang, Jinping, Paiting, and Taiyong

PQH *\ > s in Jingxian

9.0 Conclusion

This thesis has reconstructed 34 distinct onsets for Proto-Qiandong-Hmongic, the ancestral dialect of the Qianddng branch of the Hmongic family. These are shown below 108 in Figure 10.

Figure 10: PQH Onset Inventory

*p *ph *t *jh *c *ch *k *nc *nch

f ^^ ^y *s *J *Jh *9 *j *{jh *x i *ts *tsh *t9h *nts *ntsh *nt9 *nt9h *1 *1J

This may be somewhat of an artificial structure, as the lack of PQH alveolar and palatal aspirated fricatives suggest that the prenasalized alveolars and palatal forms may be older than the common ancestral forms of the labiodentals and laterals. At several points in this thesis pre-Qiandong phenomena and stages of development have been described, suggesting the branching of Qiandong away from other branches of Hmongic was not a single-step process. A more thorough ordering of changes will likely be needed from future research. Nonetheless, identifying origins at this level for modern Qiandong onsets is a significant step forward in humanity's understanding of Hmong-Mien language history. With further fieldwork and consequent reconstructions, this table and these stages of development will be refined.

The reconstruction also has paid dividends in the typological understanding of pathways of change leading to aspirated fricatives. The labiodental fricatives, the lateral fricatives, and the alveolar and palatal fricatives in Qiandong each arose in very different 109 ways, yet produced a very uniform system of aspirations, suggesting that the different changes reinforced each other in some manner. While the alveolar and palatal consonants' origins had already been probed by Wang (1979), an origin has been posited for Qiandong labiodental fricatives that until now had only been part of a Late Middle

Chinese reconstruction, and the lateral fricatives have had a source in aspirated liquids posited, which is the first case of aspirated liquids specifically, rather than other sonorants, spirantizing to create aspirated fricatives. These results both broaden and add examples to the pathways of aspirated spirant creation found by Jacques (2011).

Finally, this thesis lays the consonantal groundwork for the rimes and vocalic system to be approached for reconstruction. Ultimately, my hope in writing this is that the study of this particular branch will inform further work at the Hmongic and Hmong-

Mien levels and across the family. BIBLIOGRAPHY

Abercrombie, David. 1967. Elements of general phonetics. Edinburgh: University Press.

Andersen, Henning. 1973. Abductive and deductive change. Language 49:765-793.

Benedict, Paul K. 1986. The Na-e enigma. Linguistics of the Tibeto-Burman Area 9.1:89-90.

Benus, Stefan and Gafos, Adamantios I. 2007. Articulatory characteristics of Hungarian 'transparent' vowels. Journal of Phonetics 35:271-300.

Bloomfield, Leonard. 1933. Language. Chicago: University of Chicago Press.

Bybee, Joan. 2003. Phonology and language use. Cambridge: Cambridge University Press.

Catford, J.C. 1977. Fundamental problems in phonetics. Bloomington: Indiana University Press.

Catford, J.C. 1988. A practical introduction to phonetics. Oxford: Clarendon Press.

Chang, Kun. 1953. On the tone system of the Miao-Yao languages. Language 29-.37A- 378.

Chang, Kun. 1966. A comparative study of the Yao tone system. Language 42:303-310.

Chang, Kun. 1972. The reconstruction of Proto-Miao-Yao tones. Bulletin of the Institute of History and Philology 44.4:541-628.

Chang, Kun. 1976. Proto-Miao initials. Bulletin of the Institute of History and Philology 47.2:\55-217.

Chen, Matthew, and Wang, William S-Y. 1975. Sound change: Actuation and implementation. Language 51:255-281.

Chen Qiguang. 1984. Sheyu zai miaoyaoyu li de diwei [The place of Ho-Ne in the Hmong-Mien language family], Yuydn Ydnjiit 1:200-214.

Cho, Young-mee Yu. 1991. Voiceless sonorants are aspirates. Paper presented at the 65th Annual Meeting of the Linguistic Society of America, Chicago. Clements, G.N. 1985. The geometry of phonological features. Phonology Yearbook 2:225-252.

Downer, Gordon. 1973. Strata of Chinese loanwords in the Mien dialect of Yao. Asia Major 18.1:1-33.

Edmondson, Jerald A. 2008. Shan and other northern tier southeast Tai languages. In The Tai-Kadai languages, Anthony V.N. Diller, Jerold A. Edmondson and Luo Yongxian (eds.), 184-206. London: Routledge.

Esquirol, Joseph. 1931. Dictionnaire Kanao-Franqais et Franqais-Kanao. i-xliii, 1-519. Hong Kong.

Gleason, H.A. Jr. 1961. An introduction to descriptive linguistics. New York: Holt, Rinehart and Winston.

Halle, Morris, and Stevens, K.N. 1971. A note on laryngeal features. MIT Research Laboratory of Electronics Quarterly Progress Report 101:198-213. Cambridge: Massachusetts Institute of Technology. Reprinted in Halle, Morris (2002). From memory to speech and back: papers on phonetics and phonology 1954—2002. Berlin & New York : Mouton de Gruyter.

Haller; Felix. 2004. Dialekt und Erzahlungen von Themchen, Bonn: VGH Wissenschaftsverlag.

Haudricourt, Andre G. 1954. De l'origine des tons en vietnamien. Journal Asiatique 242:69-82. Reprinted in Problemes de phonologie diachronique, by Andre G. Haudricourt, 1987, 183-208. Paris Societe pour l'Etude des Langues Africaines=SELAF.

Haudricourt, Andre G. 1966. The limits and connections of Austroasiatic in the northeast. In Studies in Comparative Austroasiatic Linguistics 44-56. The Hague: Mouton de Gruyter.

Heffner, R-M S. 1950. General phonetics. Madison: University of Wisconsin Press.

Hock, Hans Heinrich. 1991. Principles of linguistic reconstruction. Berlin: Mouton de Gruyter.

Hockett, Charles F. 1955. A manual of phonology. Baltimore: Waverly Press. Hoenigswald, Henry M. 1960. Language Change and Linguistic Reconstruction. Chicago: University of Chicago Press.

Honeybone, Patrick. 2002. Germanic Obstruent Lenition: some mutual implications of theoretical and historical phonology. Ph.D dissertation, University of Newcastle upon Tyne.

Institute of National Languages. 1962. A brief description of the Miao language. In Zhongguo Yuwen 111.

Iverson, Gregory K. and Salmons, Joseph C. 1995. Aspiration and laryngeal representation in Germanic. Phonology 12:369-96.

Jacques, Guillaume. 2011. A panchronic study of aspirated fricatives, with evidence from Pumi. Lingua 121.9.

Jamieson, Carole. 1996. Diccionario Mazateco de Chiquihuitlan. Oaxaca, Tucson: Instituto Lingiiisco de Verano, A.C.

Janson, Tore. 1983. Sound change in perception and production. Language 59:18-34.

Johnson, Michael. 2000. The reconstruction of labial stop + sonorant cluster in Proto- Far-Western-Hmongic. Transactions of the Philological Society 100.1:25-58.

Kim, Chin-Wu. 1970. A theory of aspiration. Phonetica 21:107-116.

Kingston, John. 1990. Articulatory binding. In Kingston, John and Beckman, Mary E. (eds.), Papers in Laboratory Phonology I: Between the Grammar and Physics of Speech, 406-434. Cambridge: Cambridge University Press.

Kirk, Paul. 1966. Proto-Mazatec Phonology. PhD dissertation, University of Washington.

Klar, Kathryn A. 1977. Topics in historical Chumash grammar. Ph.D dissertation, University of California - Berkeley.

Knapp, Ring. 1996. Fonologia del Mazahua. Tesis de licenciatura, Mexico, ENAH.

Krishnamurti, Bh. (1998). Regularity of sound change through lexical diffusion: A study of s > h > 0 in Gondi dialects. Language Variation and Change 10:193—220. Kwan, Julia Chin. 1966. Phonology of a Black Miao dialect. Master's thesis, University of Washington.

Lass, Roger. 1984. Phonology: An Introduction to Basic Concepts. Cambridge: Cambridge University Press.

Lass, Roger. 1997. Historical linguistics and language change. Cambridge: Cambridge University Press.

LI YunbTng. 1995. Bahengyu zai miaoyaoyu zhong de diwei [The position of Pa-Hng in Hmong-Mien], Master's thesis, Central University of Nationalities, Beijing.

Lindeman, Fredrik Otto. 1987. Introduction to the laryngeal theory. Oslo: Norwegian University Press.

Lisker, Leigh, and Abramson, Arthur S. 1964. A cross-language study of voicing in initial stops: acoustical measurements. Word 20:384-422.

Ma, Xueliang. 1957. A script for the Miao people. China Reconstructs 6.7.

Ma, Xueliang and Tai Changhou. 1956. A preliminary survey of the phonology of the Miao dialects in the southeast of Kweichou. Yuyan Yanjiu 1:265-282. Translated in Purnell, Herbert C. (Ed.). 1972. Miao and Yao Linguistic Studies: Selected Articles in Chinese, Translated by Chang Yu-Hung and Chu Kwo-ray. Ithaca, NY: Cornell University. Pp 27-55.

Maddieson, Ian. 1984. Patterns of Sounds. Cambridge: Cambridge University Press.

Mailhammer, Robert; Restle, David; and Vennemann, Theo. In press. Preference laws in change.

Malmberg, Bertil. 1963. Phonetics. New York: Dover Publications.

Mao Zongwu and Meng Chaoji. 1982. Boluo sheyu gaishu [Short description of the Ho- Ne tongue of Boluo]. In Minzu Yuwen 6:64-80.

Matisoff, James A. 2000. Tibeto-Burman tonology in an areal context. In Shigeki Kaji, (ed.), Proceedings of the Symposium "Cross-Linguistic Studies of Tonal Phenomena: Tonogenesis, Typology, and Related Topics." Tokyo: ILCAA.

Matisoff, James A. 2003. Handbook of Proto-Tibeto-Burman: system and philosophy of Sino-Tibetan reconstruction. Berkeley & Los Angeles: University of California Press.

Mester, R. Armin and Ito Junko. 1989. Feature predictability and underspecification: Palatal prosody in Japanese mimetics. Language 65:258-293.

Michailovsky, Boyd. 1975. Some Tibeto-Burman sound changes. Papers from the First Meeting of the Berkeley Linguistics Society, 322-332. University of California - Berkeley.

Mortensen, David. 2000. Sinitic loanwords in two Hmong dialects of Southeast Asia. Ms. www.pitt.edu/~drm31

Mowrey, Richard, and Pagliuca, William. 1995. The reductive character of articulatory evolution. Rivista di Linguistica 7:37-124.

Murray, Robert W. 1982. Consonant cluster developments in Pali.. Folia Linguistica Historia 111.2: 163-184.

Nasukawa, Kuniya. 2005. A unified approach to nasality and voicing. Berlin and New York: Mouton de Gruyter.

Niederer, Barbara. 1998. Les langues Hmong-Mjen (Miao-Yao): phonologie historique. Munich: Lincom Europa.

Niederer, Barbara. 2004. Pa-Hng and the classification of the Hmong-Mien languages. In Hmong/Miao in Asia, Tapp, Michaud, Culas, and Lee (eds), 129-146.

Ogura, Mieko. 1995. The development of Middle English i and u: a reply to Labov (1992, 1994). Diachronica 12.1:31-53.

Ohala, John J. 1981. The listener as a source of sound change. In Papers from the Parasession on Language and Behavior, 178-203. Chicago: Chicago Linguistics Society.

Okell, John. 1969. A reference grammar of colloquial Burmese. London: Oxford University Press.

Oliveira, Marco Antonio de. 1991. The neogrammarian controversy revisited. International Journal of the Sociology of Language 89:93-105.

Pagliuca, William, and Mowrey, Richard. 1987. Articulatory evolution. In Papers from the Seventh International Conference on Historical Linguistics, eds. A. G. Ramat, O. Carruba and G. Bernini, 459-472. Amsterdam: John Benjamins.

Peterson, G.E., and Shoup, J.E. 1966. A physiological theory of phonetics. Journal of Speech and Hearing Research 9:5-67.

Petursson, Magnus. 1976. Aspiration et activite glottale. Phonetica 33,169-198.

Phillips, Betty. 2006. Word Frequency and Lexical Diffusion. New York: Palgrave.

Pierrehumbert, Janet B. 2001. Exemplar dynamics: Word frequency, lenition, and contrast. In J. Bybee & P. Hopper (Eds.), Frequency effects and the emergence of linguistic structure 137- 157. Amsterdam: John Benjamins.

Pike, Kenneth and Eunice Pike. 1947. Immediate constituents of Mazatec syllables. International Journal of American Linguistics 13:78-91.

Pittayaporn, Pittayawat. 2009. The phonology of Proto-Tai. PhD dissertation, Cornell University.

Pulleyblank, Edwin G. 1984. Middle Chinese, a Study in Historical Phonology. Vancouver: University of British Columbia Press.

Purnell, Herbert C. 1970. Toward a reconstruction of Proto-Miao-Yao. PhD dissertation, Cornell University.

Rankin, Robert. 1988. "FH" (the origin of the Ofo aspirated fricative). VIII annual Siouan and Caddoan Linguistics Conference, Billings,MT, June 1988.

Rankin, Robert. 2003. The comparative method. In The Handbook of Historical Linguistics, Janda, Richard D., and Joseph, Brian D., (eds.), 183-212. Oxford: Blackwell Publishing.

Rankin, Robert. 2004. The Ofo Language of Louisiana: Philological Recovery of Grammar and Typology. LAVIS III: Language Variety in the South: Historical and Contemporary Perspectives, University of Alabama. http:// www, as.ua. edu/lavi s/handouts/rankin.pdf

Ratliff, Martha. 1998. Ho-Ne (She) is Hmongic: one final argument. Linguistics of the Tibeto-Burman Area 21.2:97-109.

Ratliff, Martha. 2010. Hmong-Mien language history. Canberra: Pacific Linguistics. Rensch, Calvin R. 1976. Comparative Oto-Manguean Phonology. Bloomington: Indiana University Publications. Ringe, Donald. 2006. A Linguistic Histoty of English. Volume I: From Proto-Indo- European to Proto-Germanic. Oxford: Oxford University Press.

Sagart, Laurent. 1999. The roots of Old Chinese. Amsterdam Studies in the Theory and History of Linguistic Science 184. Amsterdam: John Benjamins.

Silverman, Daniel; Blankenship, Barbara; Kirk, Paul; and Ladefoged, Peter. 1995. Phonetic structures in Jalapa Mazatec. Anthropological Linguistics 37.1:70-88.

Sommerfelt, Alf. 1962. Diachronic and synchronic aspects of language. The Hague: Mouton de Gruyter.

Stevens, Ken. 1991. Acoustic phonetics. Ms. Massachusetts Institute of Technology.

Strecker, David. 1987. The Hmong-Mien languages. Linguistics of the Tibeto-Burman Area 10.2: 22-42.

Theraphan, L-Thongkum. 1993. A view on Proto-Mjuenic (Yao). Mon-Khmer Studies 22: 163-230.

Thomas, David. 1992. On sesquisyllabic structure. Mon-Khmer Studies 21:206-210.

Thrainsson, H. 1978. On the phonology of Icelandic preaspiration. Nordic Journal of Linguistics 1:3-54.

Vaux, Bert. 1998. The laryngeal specifications of fricatives. Linguistic Inquiry 29.3: 497- 512.

Vennemann, Theo. 2000. Triple-cluster reduction in Germanic: Etymology without sound laws? Historische Sprachforschung [Historical Linguistics] 113:239-258.

Walker, Rachel; Byrd, Dani; & Mpiranya, Fidele. 2008. An articulatory view of Kinyarwanda coronal harmony. Phonology 25:499-535. Cambridge: Cambridge University Press.

Wang FushT. 1979. Miaoyu fangyan shengyunmu bijiao [A comparison of rhymes and initials in Miao dialects], Beijing: Zhongguo shehui kexueyuan.

Wang FushT, and Mao Zongwu. 1995. Miaoydoyu gouni [Reconstruction of the sound system of Proto-Miao-Yao], Beijing: Zhongguo shehui kexue chubanshe [China Social Sciences Press],

Wang, William S-Y. 1969. Competing changes as a cause of residue. Language 45:9- 25.

Wheatley, Julian K. 2003. Burmese. In The Sino-Tibetan Languages, Thurgood, Graham and LaPolla, Randy (eds.), 195-207. London: Routledge.