Since some Sino-Tibetan languages are reported to have 'tense' vowels, one might expect those 'tense' vowels to share origins and/or distributional characteristics with the 'tense' vowels of Tangut.  But as far as I can tell, this is not the case.

The tense vowels of Hani reflect original final stops rather than lost preinitials:

Hani: *-Vp/t/k > *-Vʔ > *-Ṿʔ-Ṿ

Tangut: *sCV > *sC̣V > *C̣V > *C̣Ṿ > CṾ

(For ease of comparison with Tangut, I will use subscript dots to represent tenseness in Hani and other languages during this series of posts.)

Pre-Tangut *-k became -w in some cases and other final stops were completely lost*:

*-Vk > *-Vɣ > *-Vɰ > -V(w)


Hani tense vowels can only follow nonaspirated initials.  Conversely, Hani lax vowels can only follow aspirates and voiced initials.

initial aspirated - + -
voiced - +
vowel tense + - +
lax - + +

(This table is adapted from Sagart and Xu [2001].)

Hani seems to have a constraint against 'double laryngeal' syllables.  A Hani syllable can either have aspiration or tenseness (glottalization?), but not both.

Tangut tense vowels have no such restrictions:

initial aspirated - + -
voiced - +
vowel tense +

Next: Tenseness in Jingpho.

*4.20.1:28: Guillaume Jacques (2006) has identified one possible instance of Tangut -w from an earlier *-p:

TT5785 lwew R44 1.43 'vapor' < ?*Pʌ-lep : gDong-brgyad rGyalrong tɤ-jlɤβ < *-ɔp 'id.'

(I have reconstructed a Proto-rGyalrong labial vowel on the basis of Somang ta-jlôp.)

Perhaps pre-Tangut *-p became Tangut -w after front vowels:

Vowel type Pre-Tangut Tangut: loss of *-ɰ
Stage 1: final *-p Stage 2: lenition to final *-w Stage 3: assimilation of *-w after central vowels; loss of *-w after back vowels
front *-ip *-iw
*-ep *-ew
central *-ɨp *-ɨw *-ɨɰ
*-ap *-aw *-aɰ -a
back *-up *-uw *-u
*-op *-ow *-o

(This table deals only with the six basic pre-Tangut vowels and excludes their later 'bent' descendants.)

Note, however, that the rGyalrong word for 'vapor' has a back rounded vowel. PINE HAIR AND HARD THREAD* (PART 1)

The previous post demonstrated how vague the term 'tenseness' can be.  'Tense' vowels in Tangut and 'tense' vowels in Twi are very different phenomena.  One might expect 'tenseness' to have a single meaning in an East Asian linguistic context,.  However, as Edmondson et al. wrote (2001: 91; color version of their article here),

The field of Sino-Tibetan, Mon-Khmer, Miao-Yao, and Kadai languages is replete with statements about "tense", "lax", and other terms without the accompanying detailed anatomical and photographic descriptions of what these terms mean. Although auditory descriptions of different phonation types have been better described since Laver (1980), there have been until now no good photographic images of what is actually happening in the larynx in real language data from native-speaking informants.

Since I am not a phonetician and have no way of finding out "what is actually happening in the larynx" - particularly in the case of a dead language like Tangut - I can only look at the possible origins and the observable behavior of 'tense' vowels.  And what little I have seen so far doesn't indicate that Tangut 'tense' vowels were like 'tense' vowels in living Sino-Tibetan languages.

Next: 'Tenseness' in Hani.

*The title refers to the components of the graphs for Md 鬆緊  songjin 'lax and tense':

song 'loose'  = 髟 'hair' (semantic) + 松 song 'pine' (phonetic)

The Middle Chinese rhyme dictionary Guangyun defined 鬆 as 'disheveled (i.e., loose) hair'.

jin 'tight' = 臤 qian 'hard' (semantic/phonetic) + 糸 'thread' (semantic) 'TENSENESS' IN TANGUT AND TWI

Gong Hwang-cherng's reconstruction of Tangut has a tense vowel corresponding to every short lax vowel:

front central back
lax tense lax tense lax tense
high i ɨ ɨ̣ u
mid e ə ə̣ o
low a

I reconstruct a much more elaborate vowel system, but the one-to-one correspondence pattern generally remains the same. Three gaps appear systematic (no 'low series' ʊ [tense or lax] and no tense low central vowels and ʌ̣) and one is quite unexpected (no tense ẹ).

Tangut tense vowels may have been characterized by laryngeal tension that was conditioned by lost (acute?) preinitials:

*sCV > *C̣Ṿ > CṾ

(cf. Middle Korean sCV > modern Korean C̣Ṿ /C̣V/ in which the tenseness of the consonant and not the vowel is phonemic)

Twi is also said to have 'tense' and 'lax' vowels, but these terms refer to advanced tongue root (ATR) vs. retracted tongue root (RTR) and not to laryngeal tension. Moreover, Twi has ATR harmony whereas Tangut 'tense' and 'lax' vowels are not part of any known harmonic system. Twi only has a single low vowel, but its relative Fante has a pair of 'tense' or 'lax' low vowels:

front back
lax tense lax tense
ATR - + - +
high ɪ ~ e i
ʊ ~ o u
mid ɛ e ɔ o
low (Fante only) æ ɑ

As the chart indicates, there is some overlap between the -ATR high and +ATR mid categories, and some dialects of Twi seem to have only seven vowels:

ATR front back
tense + high i u
upper mid e o
lax - lower mid ɛ ɔ
low ɑ

Some 'Altaic' languages near the Sinosphere have vowel systems with tongue root harmony:

Khalkha Mongolian with three +/-RTR (pharyngeal/nonpharyngeal) pairs (a-e, ɔ-o, ʊ-u; see Svantesson et al. 2005 for details)

labial - +
RTR + - + -
high i (neutral)
ʊ u
nonhigh a e ɔ o

Middle Korean with three +/-ATR pairs (ə-a, ɯ-ʌ, u-o)

front central back
labial - +
+ATR i (neutral) ə ɯ u
-ATR a ʌ o

Notice that u and o have different pairings in the two languages. I've proposed that early Korean had a front +/-ATR pair *i-*e. One might be tempted to reconstruct an earlier Mongolian high front +/-RTR pair *ɪ-i by analogy with ʊ-u, but earlier Mongolian had a +/-palatal vowel system:

labial - +
palatal - + - +
high i (neutral)
u y
nonhigh a e o ø

i was phonetically palatal but neutral for harmonic purposes, implying that an earlier nonpalatal could have merged with *i.

Perhaps the depalatalization of y and ø to u and o forced original u and o into new, lower positions, causing a palatal harmony system to be reinterpreted in terms of RTR:

y > u > ʊ

ø > o > ɔ

The vowels of my 'early Old Chinese' reconstruction could be described in terms of +/-RTR instead of nonemphatic/emphatic:

front nonlabial back nonlabial [+RTR]
with central [-RTR] counterparts
back labial
RTR + - + - + -
high *i *u
mid *e *o
low *a

This OC system and its RTR harmonic rules long predate the Khalkha and MK systems. The early Mongolic language(s) in OC times might have had palatal harmony, but early Koreanic had no harmony of any sort, and MK developed ATR harmony when Mongolic had palatal harmony. Harmony systems seem to indepedently arise and fall in East Asia. WE'RE LOST: U AND I ARE MISSING IN LILLOOET

I first learned of Lillooet 15 years ago when my mentor Anatole Lyovin mentioned that he had done fieldwork on it. I never imagined that I'd ever try to find typological parallels between it and Old Chinese. I was hoping that Lillooet's 'retraction' was like OC 'emphasis'.

What initially caught my eye was the similarity between the nonretracted/retracted vowel pairs of Lilooet and my reconstructed 'early middle OC'* nonemphatic/emphatic vowels:

Lillooet nonretracted vowel Early Middle OC nonemphatic vowel Lillooet retracted vowel Early Middle OC emphatic vowel
(no high vowel!) *i (no high vowel!) ʕ
e *e ɛ ʕ
ɛ *a a ʕ
ə ʌ ʕ
o *o ɔ ʕ
(no high vowel!) *u (no high vowel!) ʕ

This chart contains two surprises. First, the nonretracted counterpart of a is phonetically identical to the retracted counterpart of e. Second, Lillooet apparently has no high vowels at all, in spite of an orthography that contains the letters u and i. It's the opposite of Proto-Indo-European, which has bothered me for a long time because it has been reconstructed without any low vowels**.

Lillooet has a third surprise: it has only four pairs of nonretracted/retracted consonants, and only one has a parallel in OC which had *l and *lʕ:

Nonretracted Retracted (velarized)
tʃ̻ (laminal) tʃ̺ɣ (apical)
ʃ ʃɣ
l lɣ
(glottalized) ɣ (glottalized and velarized!)

Why would velarization become distinctive only after these four? Or are these the last remnants of a distinction that once permeated the system (or at least its acute subset)? If so, why did these four keep what all the others had lost?

A fourth and final surprise: Retracted vowels aren't necessarily associated with postvelars, judging from the fact that the Wikipedia page contains sequences like qa [qɛ] and qe [qe] as well as qao [qa]. This is quite different from my OC reconstruction in which uvulars are followed by emphatic vowels unless they have been fronted by a preceding nonemphatic (pre)syllable.

*In "The People of the Red Ochre River", I used the terms 'early OC' and 'middle OC' without explanation. These terms refer solely to hypothetical stages in the development and loss of emphasis:

Pre-OC: No emphasis.

Early OC: Emphasis and emphatic harmony developing. Emphasis or its absence only partly reflected in the 'segregating' phonetic series of early sinography.

Early Middle OC: Emphatic vowels lowered and/or backed.

Late Middle OC: Diphthongization conditioned by emphasis or its absence.

Late OC: Emphasis disappearing; no longer needed since bent vowels are sufficient for contrast.

Late OC is more or less Schuessler's Later Han Chinese reconstruction with a few minor changes and dates from the beginning of the first millennium AD. I haven't figured out any dates for the other stages yet.

**Perhaps PIE *e and *o represented nonhigh front and back vowels which may have had low allophones. If PIE */j/ and */w/ are interpreted as vowels, the 'vowel space' of PIE could be cut up into four quadrants:

front / palatal back / labial
high *i *u
nonhigh *e [æ]? *o [ɔ]?

I wish there was an easy way to find languages in UPSID that have no high vowels (like Lillooet) or no low vowels (like PIE). THE PEOPLE OF THE RED OCHRE RIVER

speak a language that has some of the emphatic consonants I couldn't find in UPSID! Thanks to David Boxenhorn for the reminder. I had forgotten that I had found that Wikipedia article back on March 5 when I was reading about vowel-consonant harmony. It somehow slipped off my to-blog list, so I'm writing about it now before it no longer registers on my radar screen.

Chilcotin has a trio of 'flat' (corresponding to my use of 'emphatic') affricates: unaspirated tsʕ, aspirated tshʕ, and ejective ts'ʕ. The first two were also in Old Chinese.

Many OC emphatics have no Chilcotin counterparts because Chilcotin flat consonants can only be alveolar or postvelar. Moreover, Chilcotin has a three-way distinction between 'neutral', 'sharp', and 'flat', whereas OC only had a two-way distinction between nonemphatic (= 'sharp') and emphatic (= 'flat').

Chilcotin does have an binary 'nonflattened'-'flattened' distinction in its vowel system. The differences between the two types are similar to my proposed phonetic values for middle OC vowels (which varied depending on period and dialect):

Chilcotin nonflattened vowel Early OC nonemphatic vowel Middle OC nonemphatic vowel Chilcotin flattened vowel Early OC emphatic vowel Middle OC emphatic vowel
i *i *i əi ~ e *iʕ ʕ ~ *ɛʕɪʕ ~ *ʌʕɪʕ
ɪ *e *ie əɪ *eʕ ʕ ~ *ɑʕɛʕ
æ *a *ɨa a *aʕ ʕ
ɛ *ɨə ə ʕ ʕɯʕ
ʊ *o *uo ɔ *oʕ ʕ ~ *ɑʕɔʕ
u *u *u o *uʕ ʕ ~ *ɔʕʊʕ ~ *ʌʕʊʕ

(This chart excludes the OC presyllabic vowels which I write as [nonemphatic] and ʕ [emphatic].)

Note, however, that Chilcotin only flattens (lowers and backs) vowels, whereas OC bends vowels both upward (*e > *ie) and downward (*eʕ > *ɛʕ ~ *ɑʕɛʕ).

Chilcotin has both left-to-right and right-to-left flattening, and I think OC also had bending in both directions: e.g.,

left-to-right (presyllable to syllabic core)

early OC *kʕʌʕ-ki > middle OC *kʕʌʕ-kʕʌʕɪʕ

early OC*kɯ-kʕeʕ > middle OC *kɯ-kie

left-to-right (within monomorphemic or reduplicative native disyllabic words):

early OC *kʕaʕ-ku > middle OC *kʕɑʕ-kʕʌʕʊʕ

early OC *kɨ-kʕoʕ > middle OC *kɨə-kuo

right-to-left (syllabic core to preinitial):

early OC *s-kaʕ > middle OC *sʕ-kʕɑʕ

early OC *sʕ-kə > middle OC *s-kɨə

Next: A Sino-St'at'imcets comparison. LOOKING ON THE UPSID: EMPHATIC VOWELS

Jerry Norman (1994), the father of the emphatic Old Chinese hypothesis, proposed that there were three types of OC syllables:

plain (> Middle Chinese Grade III)

pharyngealized (> Middle Chinese Grades I and IV)

retroflex (> Middle Chinese Grade II)

(I think that there were two types, plain and pharyngealized, each with a retroflex subtype.)

I reconstruct pharyngealized syllables with pharyngealized ('emphatic') segments from onset to coda: e.g.,

見 'see'

Norman's OC: *ken (pharyngealized; Norman's notation does not indicate tones or their sources [in this case, *-s])

My OC: *kens [kʕaʕeʕnʕsʕ] (cf. Khmer > ae, though this vowel warping was conditioned by voiceless initials in Khmer and not emphasis)

All six OC vowels (a i u ə e o) had pharyngealized variants (aʕ iʕ uʕ əʕ eʕ oʕ). Five of those six pharyngealized vowels appear in UPSID, along with 22 others (including 8 diphthongs):

Vowel Number of languages (out of 451) Languages
eʕ 4 Archi, Hamer, Lak, Neo-Aramaic
oʕ 3 Xu!, Archi, Neo-Aramaic
ɪʕ 2 Even, Hamer
aaʕ Xu!, Neo-Aramaic
iʕ Archi, Neo-Aramaic
øʕ (mid, not upper mid) 1 Lak
ooʕ Xu!
ɐʕ Hamer
iiʕ Neo-Aramaic

The only OC pharyngealized vowel absent from the list is əʕ. However, it was lower than its plain counterpart, so it would be close if not identical to Hamer's ɐʕ.

Even, a Tungusic language, is the only East Asian language in UPSID* with pharyngealized vowels. All the others are in the Caucasus (Archi and Lak), Middle East (Neo-Aramaic), or Africa (Hamer and !Xu). Do such vowels exist outside Asia and Africa?

*I suspect that the Manchu vowel that has been romanized as ū was (or at least originated as) the lowered, pharyngealized counterpart [ʊʕ] of u. Note that UPSID lists ʊʕ as a vowel in Even. UPSID's entry for Manchu treats ū as "a higher mid back rounded vowel".

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2008 Amritavision