Here are some examples of different preinitials attached to the same root. Although I found these sets of cognates and their glosses in Gong (1988), the reconstructions are mine:

TT5505 tsia R20 1.20 'hot'

TT5499 tshia R20 1.20 < *k-tsia 'hot'

TT5498 tshwia ([tshɥa]?) R20 1.20 < *p-k-tsia 'to make hot, to heat' (adjective > causative verb)

cf. Written Tibetan tsha 'hot' < *tsa (the aspiration of voiceless obstruent initials was originally automatic, as in English)

There is also

TT3089 tsɨa R19 1.19 'hot'

which is unusual for several reasons:

- I know of no other cases of medial -ɨ- ~ -i- (Grade III/IV) alternation

- it is the only R19 syllable with a class VI (alveolar sibilant) initial

- it was placed in the Mixed Categories volume of Tangraphic Sea amidst dz-tangraphs even though it had a voiceless initial and fanqie indicating ts- instead of dz-. The initial speller for TT3089 was TT4254 tsi R11 2.10 which could not have an initial dz- since it was in the rising tone volume of Tangraphic Sea. Did TT3089 have an alternate reading with dz-?

(I have no idea why dz-tangraphs were all placed in the MC volume instead of the level and rising tone volumes.)

The zero ~ -w- alternation in 'hot'/'make hot' is parallelled in

TT0601 dʒɨe R36 2.32 < *dʒɨe-H 'cold'

TT5497 wɨe R36 1.35 < *p-dʒɨe 'to make cold' (adjective > causative verb)

whidh also contains a tonal alternation that may reflect an earlier suffix ~ zero alternation.

The pattern of *root-H ~ *p-root is also found in

TT2996 tho R51 2.42 < *tho-H 'sole of shoe'

TT2998 thwo R51 1.49 < *p-tho 'to sole' (noun > verb)

TT0311 kiiʳ R101 2.86 < *r-kii-H 'to steal; a thief'

TT1726 kwɨəəʳ R100 1.92 < *p-r-kɨəə 'to steal; a thief' (verb/noun > verb/noun)

cf. Written Tibetan rku-ba 'to steal', Old Chinese 寇 *khos 'to steal; a thief'

though these don't share the same semantic pattern as 'cold'/'to make cold' and the latter also has a vocalic alternation.

Yesterday morning, Guillaume Jacques sent me his 2009 paper 原始西夏語的前置音 "Pre-Tangut Pre-initials". It is full of ideas that parallel my own as well as notions that hadn't occurred to me: e.g., his reconstruction of pre-Tangut *r-kuu 'to steal' with a rounded vowel which nicely matches the WT and OC cognates. (But why doesn't TT0311 have a labial vowel or glide? Is it an unrelated lookalike?) I'll discuss Guillaume's paper in detail in future posts.


In my previous post, I followed Gong (1999) and reconstructed pre-Tangut *sC-clusters as the source of Tangut tense rhymes.

In Korean, both *sC- and non-*sC-clusters merged into tense initials: e.g.,

pt-, st-, pst- > tt-

I don't reconstruct clusters like pre-Tangut *pC- as sources of Tangut tense rhymes because I believe pre-Tangut non-*sC-clusters had other reflexes in Tangut:

Pre-Tangut Tangut
CV (no preinitial) CV
*pCV CwV
*sCV (and *tCV?) CṾ (tense vowel)
*rCV (< partly *tCV?), *rV CVʳ (retroflex vowel)
*kCV ChV (cf. *kC- > Ch in Korean)
*NCV (*N- = homorganic nasal) CV (C = voiced)

In my reconstruction, Tangut alternations involving medial -w-, tense voices, retroflexion, aspiration, and voicing originated as alternations between the presence or absence of preinitial prefixes in pre-Tangut.

I use the term 'preinitial' to refer to a consonant before an 'initial' consonant with or without medial glides between it and a vowel:

preinitial initial medial vowel
(C) (C) (G) V

I assume a restricted inventory of pre-initials like that of Sagart's Old Chinese reconstruction or Classical Tibetan: e.g., no affricates or voiced obstruents.

It would be strange to reconstruct *pC- and *kC- but not *tC-. (Classical Tibetan has a complete set of stop preinitials: b- d- g-.) I suspect *tC- merged with either *sC- or *rC-.

I reconstruct preinitials as prefixes only when I know of a cognate without prefixation. Otherwise, a preinitial may either be a prefix or part of a root cluster that may be a contraction of an earlier *CVC-. TROUBLED BY TTHE-NSENESS

In recent posts, I've mentioned

thɛ̣ R63 2.53 < Tangut period northwestern Chinese 大/太 *thej (phonetically [thεj]?)

which is a Chinese loanword with an unexpected tense vowel corresponding to what was presumably a lax vowel in Tangut period northwestern Chinese. Unless TPNWC also had tense vowels like Tangut, I assume the tenseness reflects a Tangut *s-prefix added to a borrowed Chinese root:

*s-thɛ > *tthɛ > *tthɛ̣ > thɛ̣

This is modelled after Korean which has nonphonemic tense vowels after tense consonants (Martin 1992: 27):

*CtV > ttV [ttṾ] (could this become a Tangut-like tṾ with a nontense consonant someday?)

But this analogy has limits. Korean has no tense aspirates like tth. No languages in UPSID have such consonants. So is my scenario probable? Just because I think I can pronounce an aspirated version of Korean tense tt- doesn't mean such a consonant (and its counterparts at other points of articulation) actually ever existed.

Moreover, Korean has no tense voiced consonants. Yet my scenario requires sound changes such as

*sdV > *ddV > *ddṾ > dṾ

*snV > *nnV > *nnṾ > nṾ

*slV > *llV > *llṾ > lṾ

*szV > *zzV > *zzṾ > zṾ

(Does any language have sz-clusters? Digraphs for single phonemes like Hungarian sz [s] and Polish sz [ʃ] don't count.)

*sdzV > *ddzV > *ddzṾ > dzṾ

(Does any language have sdz-clusters? Classical Tibetan has sts-, but not sdz-.)

Tense (or laryngealized) voiced consonants are rare in UPSID: e.g.,

dd: 1.55%

nn: 0.89%

ll: 1.11%

zz: 0%

ddz: 0%

Should I abandon the *sC-hypothesis? I can't think of anything better at the moment. MANO MAN

describes the structure of

捈 = 扌 + 余

'extend' = 'hand' (semantic; Spanish mano) + 'I' (phonetic; Persian man)

a variant of

攄 = 扌 + 慮

'extend' = 'hand' (semantic) + 'worry' (phonetic)

the subject of my last post.

I was surprised by 捈 because 余 is an Old Chinese *la-phonetic whereas 慮 is an Old Chinese *ra-phonetic which in turn contains 𧆨 = 虍/田 OC *ra (containing yet another phonetic 虍 OC *xlaʔ or *hraʔ; the initial is disputable). Normally *l- and *r-phonetics are not interchangeable. So why does the alternate spelling 捈 exist?

The earliest attestation of the graph 捈 I can find is in Shuowen (100 AD) with the meaning 臥引 'lie down and stretch' (? - note the semantic similarity to 攄 'extend'). Its reading was Late Old Chinese *dɑ, implying Early Old Chinese *la (11.11.0:15: cognate to 延 EOC *lan < ?*la-n 'extend'?)

The earliest dictionary I can find listing 捈 and 攄 as equivalents is Jiyun (1037 AD). So one might expect them to sound alike in LMC, but their readings would have been something like *du and *ʈhy. Did the use of 捈 as a simplified variant of 攄 originate in a dialect which had shifted LMC *d to *th? (Standard modern Mandarin is a descendant of such a dialect: its reading for 捈 'extend' is tu [thu].) 捈 *thu would then be phonetically close to 攄 *ʈhy.

11.11.0:22: 延 EOC *lan 'extend' could also be a cognate of 引 EOC *linʔ 'stretch' via ablaut. EXTENDING HANDY WORRIES

While writing my last entry, I was looking for a Korean Chinese character dictionary that defined 覺 as 'sleep'. Along the way, I found the Naver entry for that graph including the definition


thŏdŭk hada

'understand' (lit. 'extend-get')

As far as I know, this compound doesn't exist in standard Mandarin, Japanese, or Vietnamese. Was it made in Korea? I doubt it, since it contains a very low frequency sinograph 攄 which

- is not among the 1,800 sinographs taught in Korean schools

- is #7,522 in this list of 13,053 sinographs ordered by frequency

- is not in this list of 9,868 sinographs ordered by frequency or Lin Yutang's dictionary.

I wonder if 攄得 thŏdŭk 'understand' was borrowed from colloquial northeastern Middle Chinese (like ?生覺 saenggak 'think'?).

The Sino-Korean reading of 攄 is unusual. I would expect 처 chhŏ < 텨 thyŏ < late MC *ʈhyə, as reflected in the prescriptive Sino-Korean reading thjə in Tongguk Chŏng'un 6: 87. But 터 thŏ implies (early?) MC *th(ɨ)ə or *ʈh(ɨ)ə. Could 터 thŏ be borrowed from early MC instead of late MC? Or does it reflect a colloquial NW MC word for 'extend' without *j?

攄 also has an unusual reading in standard Mandarin: shu instead of the expected chu. (Both the 'Phags-pa and Zhongyuan yinyun varieties of Yuan Dynasty Mandarin had *tʂhy.) I suspect sh- and ch- reflect different Old Chinese prefixes (in bold):

Modern standard Mandarin shu [ʂu] < late MC *ʂyə < early MC *ʂɨə < OC *s-(h)ra

Yuan Dynasty *tʂhy < late MC *ʈhyə < early MC ʈhɨə < OC *t-hra

攄 has a rare Sino-Vietnamese reading implying late MC *lyə < OC *ra (without any prefixes) as well as the expected SV sư, but the SV l-reading is probably due to analogy with the SV reading of its phonetic 慮 lự 'worry'. (The left-hand element is 扌 'hand'.)

Next: Mano man.

11.10.1:55: nomfoundation.org lists these three nom readings of 攄:

lựa (implying an extinct SV reading based on early MC *lɨə < OC *ra)

so (implying an extinct SV reading based on early MC *ʂɔ < OC *s-(h)ra with emphasis and irregular vowel development)

thư (implying an extinct SV reading based on late MC *thɨ < OC *t-hra)

The last reading is not far from the unusual Sino-Korean reading thŏ [thə]. STUDYING SIGHT AT NIGHT

In last night's post, I mentioned

Middle Chinese *kæwk

'awake; apprehend, get insight; rouse into understanding; (Buddhist) enlightenment'

= 學 'study' (phonetic) + 見 'see' (semantic)

as a possible source of the second half of Korean


saenggak 'think'

I did not mention its other Middle Chinese reading *kæwh which was borrowed into Korean as 교 kyo. That second reading, though rare in Korean, developed into modern standard Mandarin jiào 'sleep'(!).

I was surprised to find that the Guangyun (1008 AD) definition of this second reading was 睡覺 which is now read as MSM shuìjiào meaning 'sleep'. Did it mean 'sleep' in the 11th century?

Although 'sleep' is absent from the Korean and Japanese character dictionaries - implying that meaning didn't yet exist during the Tang Dynasty when those two languages imported most of their Chinese vocabulary - it can be found in Cantonese gaau 'sleep' and Taiwanese Hakka gau 'sleep'. So 'sleep' cannot be a late northern innovation.

Some languages geographically between Cantonese and Beijing-based MSM have different words for 'sleep'. I wish I had my copy of 汉语方言词汇 with me, but I don't, so the forms below are drawn from eastling.org and the ROC's Taiwanese and Hakka dictionaries.

Root Language Form
困/睏 Nonstandard Mandarin kun
松陽 Songyang Wu khuɛ̃
蒲門 Pumen Wu khuan
龍泉 Longquan Wu khuʌ
泰順蠻講 Taishun Manjiang khuəŋ
Taiwanese khun

Taishun Manjiang also has a noun 覺 ku 'sleep'.

Characters are not necessarily the best guides to etymology:

睏 is cognate to Cantonese fan < *khwan < *khun 'sleep', written with a different character 瞓.

Taiwanese Hakka 睏 represents an unrelated morpheme kun 'short time period'.

Cantonese 唔 represents an unrelated Cantonese word m 'not'.

睏 is not in Old Chinese or in Guangyun. Nevertheless, its wide distribution implies that it might be an old colloquial word that was not written until recent times.

I wonder what the distribution of 覺 'sleep' is/was. I assume some languages with 困/睏 once also had 覺. LIke Cantonese, Shanghai has both:

捆搞 khuəŋgo (note the graph for the first syllable - it's not 睏)

覺覺 kɔkɔ

I can't find either 困/睏 or 覺 for 'sleep' in Vietnamese. I can't even find the reading Sino-Vietnamese giáo for 覺. Vietnamese may have been heavily influenced by a Cantonese-like variety of early Chinese, but it may not have borrowed the *khun underlying Cantonese 瞓 fan or the *kjaaw underlying Cantonese gaau (which should theoretically correspond to SV giáo).

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2009 Amritavision