LINGUIST List 10.1148

Thu Jul 29 1999

Review: Wannemacher: Aspects of Zaiwa Prosody

Editor for this issue: Andrew Carnie <>

What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Andrew Carnie at


  1. Jakob Dempsey, book review

Message 1: book review

Date: Sun, 18 Jul 1999 19:35:01 +0800
From: Jakob Dempsey <>
Subject: book review

Wannemacher, Mark W. 1998 Aspects of Zaiwa Prosody: an 
Autosegmental Account. Summer Institute of Linguistics/University of 
Texas at Arlington, viii,160 pp. $29 International Academic 
Bookstore (SIL) 972-708-7404 / 

reviewed by Jakob Dempsey, Yuan-ze University 

 Modern linguistics is a collection of many disciplines, but it 
still owes much to its origins in comparative and descriptive 
studies centered on the great Indo-European language-family. The 
only other language-group of comparable size, both in number of 
speakers and certainly in number of languages, is the Sino-Tibetan 
group, but modern linguistic studies are still in the beginning 
stages for many of the hundreds of languages in this group. Zaiwa is 
a rather typical representative of the Tibeto-Burman(TB) super-
group, belonging within the relatively well-studied Yipo-Burmic 
group to the Northern Burmish sub-group, thus with close affinities 
to Burmese. Although Zaiwa is the best-documented among the known 
members of its sub-group (including Maru [=Longwo/Langsu], Bola, 
Letsi and the Achang dialects), Wannemacher's volume is its first 
detailed phonological study. Of added interest to many readers is 
the autosegmental analysis and feature-geometry which W. 
(Wannemacher) uses to explain the data he has gathered.

 The book starts with an introduction to Tibeto- Burman languages 
and Zaiwa's position in that large group, with family-tree diagrams 
from 4 different authorities; I find it strange that our author 
relies most on the oldest version (Shafer) constructed with the 
least data available. Among the general features of TB languages, he 
mentions phonemic tones and voice quality contrast, "two 
phonological traits that apparently developed independently." I wish 
he had explained this more, since I have often seen a close 
relationship between the two in both the Sinitic and TB side of 
Sino-Tibetan. After an outline of W's methodology and focus of the 
research in the following chapters, we come to chapter 2, devoted to 
previous studies of Zaiwa phonology. Syllable structure, segment 
inventory, tones, voice-quality and the distributions of all the 
above are compared from Burling 1967, Xu and Xu 1984, Yabu 1988 and 
Dai Qing-xia 1988, 1990. Since W's material, like Yabu's, is based 
on speakers from Burma, I think it would have been advisable to also 
consult Yabu's much larger dictionary of Zaiwa (1982). Also, a great 
deal of comparative material, not altogether irrelevant to W's 
tasks, can be found in the massive Huang 1992. This chapter 
naturally includes some discussion about choices made by the various 
researchers on how to phonemicise Zaiwa's sounds and includes some 
interesting details on allophonic variants. W. distinguishes himself 
from some earlier analysts by positing a single derived tone for 
checked syllables (stop-final) with tone height depending on 
features of the initial consonant and voice quality. He treats the 
latter as a suprasegmental feature, which is quite a point of 
contention among analysts. The issue here is a marked type of voice 
quality, sometimes described as "creaky" or "tense", described in 
chapter 3 onward as [+stiff vocal folds], which has been attributed 
to "glottalised initials" (Burling) or simply to the presence of a 
glottalised/tense vowel (others). In a more up-to-date analysis, W. 
assigns the tense voice quality to the initial consonant as an 
autosegment which spreads to the following vowel. As more of a 
comparative/historical linguist, the question for this reviewer 
remains, "Where did this marked feature come from?" The answer seems 
to be: not from "glottalised initials" which was just an early 
theory of Burling's with little evidence to back it up, but from 
earlier clusters of /s/ + stop that are no longer acceptable in the 
syllabic canon. A similar development is confirmed by older 
spellings of "s+C" in syllables that now start with such "tense" 
stops in Korean. In his well-ordered "Allophonic statements" tables 
(p.142-4), we can see many examples of this three-way distinction 
with stop-initial words. In generalised symbols, we have, by W's 
analysis: ga - kha - gA , representing respectively non-aspirated 
and aspirated with modal (unmarked) voicing, and tense voicing 
(always unaspirated). The apparently voiced stops are also favored 
for syllable-final position, thus: gag - khag -gAg . It may appear 
petty to quarrel about symbols used, but this is after all an 
important part of phonemic analysis as well as the very shape of the 
language as we see it presented to the world, so I offer here some 
suggestions: Looking at the trio ga - kha - gA (capital letter = 
tense voicing; W. uses a tiny "+" appended under the affected 
vowel), it is obvious that the aspiration-mark of the second member 
is non-contrastive. Since the "native script" for Zaiwa (such as 
seen in its large Chinese-to-Zaiwa dictionary) represents W's "kha" 
by "ka" , let us do that here. I do not quarrel with the /g/ itself; 
although the initial in /ga/ is only marginally voiced, it has 
served this contrastive role in many analyses, including Mandarin 
Chinese's Pinyin, where the initials are even less voiced. The 
Chinese and Japanese analysts listed above would disagree, using ka 
- kha - kA instead. As for the tense-voiced member, the choice is to 
associate the extra marked feature with g- or with k- . W. chose 
the former, and we will examine his reasoning when discussing 
chapter five. I would suggest: ga - ka - xka , where the /x/ is not 
so vital, simply being my historical bias in using a segment ( = W's 
/h/ ) to mark what W. rightly calls a suprasegmental feature in 
present-day Zaiwa. His use of /-g/ etc. for final stops is less 
pardonable since this gives us forms which are typolo- gically quite 
out of place in East Asia; there is not the least bit of voicing in 
final position, so why not use the traditional -k etc. since 
everybody knows this does not imply any aspiration finally?

 Chapter 3 "Phonological Prerequisites to the Analysis and 
Representation of Zaiwa Phonology" includes brief but well-focused 
discussions of autosegmental phonology, feature geometry and 
underspecification theory, with a focus on the laryngeal node as 
well as a detailed look at the anatomy of the larynx and its 
relationship to the two most important suprasegmentals in Zaiwa: 
contrastive pitch and voice quality. Drawing on the work of previous 
TB scholars as well as Laver and Ladefoged, W. brings out such 
points as: "...Zaiwa checked syllables in which the glottal 
constriction of the final stop and the [+/- spread vf ] and [+/- 
const vf ] features of the initial consonant work together to 
determine tone," and that creaky voice [+ const vf ] [- stiff vf ] 
has the lowest tone, with modal voiced segments [- const vf ] [- 
stiff vf ] tone is determined by surrounding segments and 
independent laryngeal controls, and tense voice segments [+ stiff vf 
] have a tendency toward higher tone.

 Chapter 4 "An Autosegmental Framework for the Analysis and 
Representation of Tone and Voice Quality in Zaiwa" early on points 
out that "In Zaiwa unchecked syllables, tone is assigned by 
associating lexically specified tone from an autosegmental tier to a 
TBU as expected. In checked syllables, however, tone is derived by 
spreading previously associated glottal features to the vowel from 
the surrounding segments." Three lexical tones (high-falling, high-
mid level, mid-low falling) are contrastive in non-checked syllables 
irrespective of initial voicing and voice quality, but tone in 
checked syllables is derived and predictable from initial (C1) 
voicing and [+/- stiff vf ]. The laryngeal feature systems of Halle 
and Stevens, Duanmu, Ladefoged and Yip are compared, whereupon W. 
suggests an analysis using [ +/- spread ], [+/- stiff ], [+/- const 
], and [+/- voice ] which can be simplified through 
underspecification. There follow a number of feature-geometry 
diagrams which were quite helpful to a reader like me with a better 
background in Zaiwa than in modern theory.

 Chapter 5, "Aspects of Zaiwa Segmental Phonology", opens with a 
list of five different phonological levels (paragraph, breath-group, 
word, syllable, phoneme) each with its own features; the syllable is 
taken as the basic frame of reference. Nuclear words may be from 1 
to 4 syllables, but verbs are as a rule monosyllabic (a common 
feature in TB), whereas nouns are commonly of either one or two 
syllables. Syllable structure is: C1(C2)V(X) [numerals should be 
subscripts] where X is optional vowel length or the second member of 
a diphthong or a final consonant. Oddly, I feel, W. considers the 
initial consonant to be obligatory, there- fore he must write a non-
contrastive glottal-stop before all otherwise initial vowels. His 
reasons (p.68-69) as I understand them are: 1) /?/ (glottal stop) is 
already a recognised segment (it can occupy the X-position) , 2) the 
sound [?] can be heard before syllable-initial V in cases such as 
V(C) and CV-V , 3) "glottal stop patterns with other stops syllable 
finally and is contrastive with its absence syllable finally," but 
nothing in any of these three statements convinces this reviewer 
that the [?] heard initially is therefore a phoneme. W. further 
argues that excluding /?/ initially would increase the number of 
syllable-types, thereby making analysis more complex. Yet on the 
same page, in a discussion of unstressed final syllables, he has to 
add a special extra rule to account for cases where the syllable 
begins with his /?/. So, his (I feel) superfluous initial /?/ can 
itself "make the analysis more complex", and in any case a 
linguist usually hopes that the transcription he/she has devised 
might by worthy of use by many other people, who in this case would 
all have to add this extra letter whether they had any interest in 
W's analyses or not.

 That said, we move on to note W's use of a moraic approach to 
syllable structure. W. connects stress with heavy syllables ( -VV or 
-VC) and nonstress with light syllables (-V), thus a Zaiwa syllable 
has at most two moras. The first mora must be of equal or greater 
sonority than the second. W. presents some insights gained from the 
traditional Chinese method of syllable-analysis, and then discusses 
reduced syllables. In the next section on consonants, W. first 
states that it is not obvious whether to assign the tense-voice-
associated initial stops to his "voiced plosive" or to his "voicless 
aspirated plosive" category, but then decides on the voiced series, 
but with reasons I find both curious and inadequate. Curious is that 
he first, rather inexplicitly, decides that tense voicing is an 
inherent quality of such syllables, with no need to search for any 
explanation of how this marked suprasegmental came about; W. earlier 
(p.33) quoted other scholars voicing this opinion (although Burling 
calls it in origin "preglottalised" rather than "tense") without 
really taking a stance, but here he has. This marked voice quality 
then would influence the underlyingly voiced initial and render it 
audibly voiceless. He then refers to some causative/non-causative 
verbal pairs such as / bup / 'rotten' vs. / bUp / 'make rot' to 
demonstrate that the causative member is basically a derivative of 
the other voiced-initial member. He has found no evidence of such 
variance with the voiceless aspirated series, but he may have 
overlooked a most pertinent article (Dai 1981) which specifically 
discusses causative verbs in Zaiwa. Dai divides the mechanisms into 
two major categories, analytical (adding the prefix / lO? / ) and 
inflexional . The latter he divides into three types: ( non-caus. 
vs. caus., using W's notation): a) ga - gA , na - nA [ g = any 
voiced stop, n = any resonant] b) ga - Kha c) special. The last 
type involves changes such as wun (carry on the back) - hun (caus.) 
; the former is from *run, the latter from *hrun (=*xrun), cf. 
Dempsey 1995 p.284. This third type is the least common, and the 
first type is the most comon. Now, in other languages within the 
"Zaiwa-group", e.g. Maru, we see quite similar mechanisms, but in 
the other North Burmish sub-type, Achang, and in Brumese, we see 
something very interesting: for the Zaiwa type ga - gA we often 
see ga - kha (e.g. 'ascend, fill, fry, adhere, fear) , and for the 
Zaiwa type na - nA we often see na - hna (e.g. red, be at, roll, 
weep, release, sink). If we assume the Zaiwa tense syllable earlier 
had an / s / prefix, since s + C is well-known as a means for 
deaspirating C, it would hardly do to maintain that *sC led to Ch 
(aspirated C). For the resonant initials, it looks very simple: *sna 
> hna ( > nA ). For the stops, it is not clear whether the tense 
type in Zaiwa corresponding to the aspirated type in Achang/Burmese 
is due to an extra /s-/ having been added on to an already causative 
aspirated type, or simply added to devoice the original voiced non-
causative. The former possibility is strengthened by looking at some 
cognate sets in North Burmish, which show the same relationship, 
tense voicing in Zaiwa but aspiration in another closely related 
language. For example, 'star' has tense voicing in Zaiwa, Maru, Bola 
and Letsi, but aspiration in Achang. 'Gall-bladder', 'cheek', 'frog' 
and 'mosquito' are other examples. In these cases, we can hardly 
suppose that the aspirated forms in Achang (and sometimes Burmese) 
are some type of morphologicall alteration like we saw with the 
causative verbs. It is simpler to assume that the Zaiwa-group forms 
show evidence of having added the typical TB body-parts /s-/ prefix 
or the common /s-/ animal prefix, cf. Benedict 1972 p.106. 'Frog' 
for example shows random aspiration throughout the Yipo-Burmic 
super-group; some languages kept or added the /s-/, some didn't. W. 
also argues a little for associating tenseness with the voiceless 
aspirated series, but with little force too since the correspon- 
dences in Hani/Akha he alludes to are allophonic. He concludes "the 
deaspiration of a not supported by the feature geometry 
proposed for Zaiwa" in which aspirated consonants are [ +spread vf 
], incompatible with tense voice's [ -spread vf ] , but the role of 
*s- is unclear to me, because on p.86 he lumps voiceless fricatives 
in with voiceless aspirates as both being [ +spread vf ] but in his 
important chart on p. 51 and in his appendix the fricatives /s/ and 
/$/ (English "sh") are listed as [ -spread vf ]. I would conclude 
the same way I did for final stops: if initial tense stops sound 
voiceless, why not write them voiceless?

 Chapter 6 "Zaiwa Suprasegmental Phonology" starts with details 
about tonal allo- phony, with an interesting distinction between 
nonproductive vs. productive reduced syllables, the former 
completely losing their underlying tone. The section on tone sandhi, 
complete with CECIL tracings, is one of the most thorough I have 
seen for a TB language. W. then reviews arguments for considering 
the high and low checked-syllables as basically one or two tone 
types; he opts for the former, but it is a pity he did not give some 
examples of the evidence from Maru and Lashi (= Letsi) which is 
involved. Sections on stress vs. tone and voice quality conclude the 

 The book has lengthy appendices which list further details of 
Zaiwa's phonological processes as well as distinctive feature 
charts, and references. One last detail bothers me: There are some 
central vowels (mid or close-mid in the Chinese sources, although W. 
hears one of them as higher) in Zaiwa which Chinese sources consider 
as allophones of /e/ and /i/ in various environments; W. does not 
hear the variant for /e/, thus /se/ (level-tone) 'know' is a front 
vowel for him, but a central vowel in the Chinese sources. Be that 
as it may, W. argues that the higher central vowel should be an 
allophone of /u/, not of /i/. I disagree with both: the Chinese 
publications seem to be based on analogous allophony in Jingpo, and 
W. takes his inspiration from Matisoff's analysis of Lahu. I instead 
look at cognates: 'Shoot', for example, which W. writes /bug/ and 
Xu&Xu write /pik/, I write as /bek/, because: a) the cognates in 
Zaiwa-group languages all have mid or low central vowels, and 
closely related Nusu has /beq/. No TB cognate I know of has /u/ 
here. b) the two different central vowels of the Chinese sources 
are, according to my investigations, actually in complementary 
distribution, so, since they are both described as mid, and since 
cognates frequently have /e,a,/ ( as in RP "love"), the mid vowel 
/e/ is a sensible home for them. c) final /-k/ (W's -g) is 
associated with front vowels in Zaiwa; *-k with back vowels has 
changed to a glottal stop. 

Benedict, Paul K. 1972. Sino-Tibetan: a conspectus. Cambridge: 
University Press. 

Burling, Robbins. 1967. Proto-Lolo Burmese. International Journal of 
American Linguistics 33(2):part 2. Bloomington: Indiana 

Dai, Qing-xia. 1981. "Zai-wa-yu shi-dong fan-chou di xing-tai bian-
hua" (Morphological changes in the Zaiwa causative-verb category), 
in Min-zu yu-wen 1981.4:36-41.

 ---------- 1986. Zaiwa-yu (the Atsi language). Zhong-guo da-bai-ke 
quan-shu: Min-zu. (Magna Encyclopedia Sinica: Ethnology Volume). 

- -------- 1993. A genetic classification for Tibeto-Burman 
languages in China. in: Recent contributions to Tibeto-Burman 
studies. Beijing: CUN Press. 

Dempsey, Jakob. 1995. A reconsideration of some phonological issues 
involved in reconstructing Sino-Tibetan numerals. diss., 
U.Washington, Seattle. 

Huang, Bufan (ed.). 1992. Zang-mian yu-zu yu-yan ci-hui ( a Tibeto-
Burman Lexicon). Beijing: Zhong-yang min-zu xue-yuan. 

Xu Xi-jian & Xu Gui-zhen. 1984. Jing-po-zu yu-yan jian-zhi (Zai-wa 
yu). (Outline of the Zaiwa language of the Kachin nationality). 
Outline of China's Minority Language Series. Beijing: 
Nationalities Publishing House. 

Yabu, Shiro. 1982. A Classified dictionary of the Atsi or Zaiwa 
language (Sadon dialect) with Atsi, Japanese and English indexes. 
Tokyo: Institute for the Study of Languages and Cultures of Asia 
and Africa, Gaikokugo Daigaku. 

- -------- 1988. A preliminary report on the study of the Maru, 
Lashi and Atsi languages of Burma. In Yoshiaki Ishizawa (ed.), 
Historical and cultural studies in Burma, 65-132. Tokyo: Institute 
of Asian Studies, Sophia University. 

Jakob Dempsey was a medical technologist for many years, but now is 
Assistant Professor in the Department of Foreign Languages at Yuan-
ze University in Taiwan, where he specializes in introduction to 
phonology and German courses. His focus of research is Old Chinese 
phonological reconstruction and Tibeto-Burman compar- ative and 
historical phonology, with occasional forays into Germanics and 
other Indo-European groups. He lived for three years in Afghanistan 
and Korea and has an M.A. in Tibetan, Ph.D. in Asian Linguistics 
(both U.Washington). He is currently finishing a paper on the 
phonological history of the North Burmish group. 

Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue