11.9.24.23:59: WHY ARE RHINOS LATE?

Until I read "Rhinoceroses in ancient China" at Wikipedia tonight, I never gave any thought to why Chinese had not just one, but two basic native words* for 'rhinoceros':

犀 Mandarin xi < Middle Chinese *sej

Vietnamese  犀 tê < *se 'rhino' is also from MC *sej

V  犀角 tê giác 'rhino' is literally 'rhino horn': i.e., 'horned rhino'. (Vietnamese has modified-modifier order.)

Japanese  犀 sai 'rhino' is from southern Middle Chinese *saj

兕 Mandarin si < Middle Chinese *ziʔ

Could these two words share a common root? I could derive them both from *Nʌ-si in my Old Chinese reconstruction:

犀 MC *sej < OC *si < *Nʌ-si < *Nʌ-si

presyllable *Nʌ- conditioned emphasis (indicated with underlining) before being lost; emphatic *i warped into *ej

兕 MC *ziʔ < OC *Nziʔ  < *N-si-ʔ < *Nʌ-si-ʔ

presyllable reduced to *N-; without low vowel to condition emphasis, *si remained nonemphatic and its vowel remained high; *N-s- fused into *z-

what is the function of the final glottal stop suffix?

However, 犀 is also thought to be phonetic in  遲 Middle Chinese *ɖi 'late'. This would not be possible if 犀 had *s- since MC *ɖ- does not derive from *s- or *s-clusters.

One workaround is to reconstruct 犀 and 兕 with *l-roots:

犀 OC *sʌ-li

遲 MC *ɖi 'late' could be from OC *rli sharing *li with 犀

兕 OC *sliʔ < *sʌ-li-ʔ

The *sʌ-l- of 兕 could have fused into *sl- and then *z-. Later, the *sʌ-l- of 犀 could have fused into *sl- which simplified to *s- (since the *sl- > *z- rule was no longer applying). In short, the same presyllable-root-initial sequence (*sʌ-l-) could have different reflexes depending on what fusion rules were applicable at the time.

Baxter and Sagart (2011) reconstructed 犀 with *l but reconstruct 兕 with uvular *ɢ, presumably by analogy with instances of MC *z- in phonetic series that they reconstruct with uvulars**:

犀 OC *s.lˤəj

兕 OC *[s.ɢ]ijʔ (the brackets indicate that they are not certain about the initial)

However, Schuessler (2007: 479, 523; 2009: 279, 280, 283) had a very different view on these two words.

First, he pointed out that 遲 'late' has a variant 遟 with the phonetic 屖 MC *si 'to wait'. He believed that 屖 is cognate to 栖 MC *sej 'keep still'. I presume that he regarded 遲 'late' as a semantic compound

辶 'to go over' + 屖 'wait' (later written as 犀 'rhino')

since MC *ɖ- cannot derive from *s- or *s-clusters in his reconstruction.

He reconstructed 犀 as OC and related it to Written Tibetan bse 'rhino' and various Sino-Tibetan words for 'elephant', 'cattle', 'deer', etc. and even Mon-Khmer words for 'horse'. I think a connection with bse is plausible (and would rule out B&S2011's reconstruction with *-l-) but am skeptical of the other proposed cognates because of their semantics..

Second, he glossed 兕 as 'wild water buffalo', specifying "not 'rhinoceros'" and derives it from Sino-Tibetan *s-jəl, the source of Lushai sial 'domestic buffalo'. He regards as an "area word" with cognates in other language families like Proto-Northern Tai *ǰɨa 'ox, cow' and even words in the Munda subgroup of Austroasiatic: Mundari sahil, saili 'wild buffalo' and Gutob saail 'wild buffalo, deer' (with an l! - cf. my *sʌ-li-ʔ above): "However, the contact [of Austroasiatic Munda in India?] with OC would be so remote that a regular phonological history is elusive."

*English rhinoceros is neither native nor basic since it's a compound of Greek roots for 'nose' and 'horn'.

Is Korean khoppulso 'rhino' (lit. 'nose-horn-ox') a calque of rhinoceros? Korean has another word muso from Middle Korean mɯsyo '?-ox' whose first element may have once been mɯr 'water'. Did *mɯr-syo originally refer to water buffalo?

**9.25.00:48: Here's an example of a B&S2011 phonetic series with MC *z- < OC *s.ɢ-. (The MC notation sans asterisks is theirs.)

訟 MC zjongH < OC *s.[ɢ]oŋ-s 'to litigate'

瓮 MC uwngH < OC *qˁoŋ-s 'earthen jar'

公 MC kuwng < OC *C.qˁoŋ 'public'

According to B&S2011, OC *q(ˁ)- became MC ʔ- unless preceded by a preinitial *C. MC *ʔ- is unwritten in their notation.

9.25.1:11: ADDENDUM: I forgot to write about the forms of the graphs 犀 and 兕:

犀 has 牛 'ox' on the bottom, but what's on top? It looks like 尸 'corpse' and five extra lines - four horizontal and one vertical - but I doubt that's what it is. The Taiwanese variant dictionary lists the top half as a variant of 'rhino'.Is it a drawing of a rhino?

兕 looks like 凹 'concave' atop a 儿 pair of legs but is presumably a distorted drawing of an animal. 兕 has many variants including one in which 四 'four' might be phonetic as well as a similar-looking substitute for 凹. (四 and 兕 have been homophonous in some varieties of Chinese for about a millennium.)

11.9.23.23:59: THE FALSE GREATNESS OF SUMMER

While reading Jerold A. Edmondson's "The power of language over the past: Tai settlement and Tai linguistics in southern China and northern Vietnam" on Wednesday, I realized that

- the syllables of 華夏 Huaxia nearly rhymed, perhaps indicating a reduplicative origin

- 夏 'great' might be cognate to 假 'great'.

華 *N-qwhˤra 'flower' (v.), 'flowery' (adj.)

華 or 花 *qwhˤra 'flower' (n.)

華 *[ɢ]wˤra-s (name of a mountain)

夏 *[ɢ]ˤraʔ 'great'

cf. 夏 *[g]ˤraʔ(-s) 'summer' (cognate - the great season?)

假 *kˤraʔ 'great'

cf. 假 *Cə.kˤraʔ 'borrow; false' (unrelated homophone written with the same graph)

I don't understand why B&S2011 has

- different initials for 夏 'great' ~ 'summer'; mixed uvular-velar-initial phonetic series are unusual

- different points of articulation for the initials of 夏 'great' and 假 'great' which prevent me from regarding them as cognates

Here are two alternate reconstructions within the B&S2011 framework as I understand it. Differences from B&S2011 are in bold.

Solution A involves only one change. 夏 'great' is a prefixed variant of 假 'great'.

Solution B is more costly. All the initials have been changed. If 夏 'great' had a uvular root initial, then its near-homophone 夏 'summer' and its cognate 假 'great' must also have had uvular root initials. The problem is that neither 夏 nor 假 belong to phonetic series with Middle Chinese *ʔ-,*x-, or *j- initials that point to Old Chinese uvulars in the B&S2011 system. (The other initials are not diagnostic because they are found in both OC uvular and velar series.)

Middle Chinese reflexes of B&S2011 uvulars

(The four initials in the "without OC *C.-" column match the Tangut class VIII initials! Coincidental, or ... ?)

Both 夏 and/or 假 could have had uvulars, but making them uvular simply because B&S2011 reconstructed 'great' as *[ɢ]ˤraʔ with a potential uvular (why?) is too drastic. Nonetheless, let's take that as a given for now and see far we can go.

華夏 *N-qwhˤra-*N-qˤraʔ could nearly fit the pattern *NQaNQa. Phonetic similarity is a big plus for a catchy phrase.

Could 華 *N-qwhˤra 'flowery' and 夏 *N-qˤraʔ ~ 假 C.qˤraʔ 'great' be cognates? Although I can imagine a semantic shift

'flower' (n.) > 'flowery' (adj.) > 'great'

I don't think that's what happened here because 'flower' has an aspirated labiouvular root initial, whereas 'great' has an unaspirated uvular root initial. If one tries to reconcile the two by positing prefixes to account for the aspiration and labial coarticulation

*S-P-qˤra > *qwhˤra

one ends up with *(N-/C.)qˤraʔ 'great' having a simpler form than its supposed source 'flower'. I would expect more affixation in derivatives than in their sources.

9.24.00:27: 華 is usually reconstructed with an *-r- because its Middle Chinese reading *ɣwæ has a low front vowel that is normally a reflex of OC *ra. However, Schuessler (2009: 51) pointed out the curious fact that there are no syllables of the type *wâ (= B&S2011's *ɢʷˤa) and proposed that 華 might not have had an *-r in OC:

OC *wâ = *ɢʷˤa > MC *ɣwæ

Did a low back vowel *[ɑˁ] (or even rounded *[ɒˁ]?) disassimilate to [æ] after *wˁ?

11.9.22.23:57: <Λ>NOMALOUS VOWEL SYSTEMS

In my last entry, I looked at T, inverted-T, and I-shaped vowel systems. I might as well look for vowel systems shaped like the rest of the alphabet and then some.

In theory, if one were to program a computer to generate a vowel system, one could give it instructions like:

- pick a random number of vowels from, say, 2 to 9

- fill in the following nine slots with those vowels at random:

But in reality, vowels are not distributed at random. I don't know if some distributional patterns exist at all.

1. Λ (lambda)-shaped

is the opposite of the 'vowel triangle' found in Classical Arabic:

The Λ shape violates what seems to be a universal: a vowel system must have at least one low vowel.

The only counterexample I can think of is Proto-Indo-European.

UPSID lists Mari as a language without low vowels in spite of the a in its name and the letter а in its variant of the Cyrillic alphabet. Is this a mistake, or does UPSID's lower mid back unrounded [ʌ] for Mari corresponds to the Mari letter а?

This violates the same universal as the lambda shape:

The Proto-Austronesian vowel system has an ∀-shape:

A similar ∀-shaped-system was reconstructed for pre-Old Japanese (Martin 1987) and Old Chinese (Li Fang-Kuei 1971 and Gong Hwang-cherng 1995). However, nowadays a Џ (Cyrillic dž)-shaped system is reconstructed for those languages:

I also think pre-Tangut had a Џ-shaped system which expanded into a much larger system due to the incorporation of compression artifacts: i.e., remnants of lost segments.

4. > (greater-than)-shaped

5. Diamond or plus-shaped

6. L-shaped

7. Reverse-L-shaped

These five shapes go against a tendency for vowel systems to have filled upper corners. Their a-less variants may violate the a-universal I proposed unless they have front or back low vowels (e.g., æ or ɑ).

The rarity (or absence?) of shapes 3-7 may lead one to think that only a subset of symmetrical vowel systems are possible. Not so.

A case of actual asymmetry: Manchu

Last night I found Kiyose's (1998) "Dialectal Lineage from Jurchen to Manchu" in which he reconstructed the Jin Jurchen vowel system as

which was then reduced to the five-vowel system of Ming Jurchen

that does not match the six-vowel system of Manchu which I interpret as

Manchu ə and ʊ are traditionally romanized as e and ū.

It is unlikely that Ming Jurchen u split into two Manchu phonemes u and ʊ.

The mismatch between the Jurchen and Manchu systems may indicate that the two languages are 'aunt' and 'niece' rather than 'mother' and 'daughter': i.e., Manchu is not a direct descendant of Jurchen but is a descendant of a sister of Jurchen.

Another possibility is that the Chinese transcribing Jurchen - and Jurchen writing in the Jurchen script - overlooked the u : ʊ distinction recorded in the later Manchu alphabet. Absence of evidence for ʊ is not necessarily evidence of absence. (9.23.00:08: One would not deny that Old Mandarin had tones simply because tones were not recorded in the Hphags-pa script.)

In any case, the Manchu system is not unlike that of Middle Korean:

Both Manchu and Middle Korean have more back vowels than front vowels.

I wonder if there are languages with mirror-images of their systems: e.g.,

with four front vowels but only one back vowel.

11.9.21.23:57: DID PROTO-HLAI HAVE A T-SHAPED VOWEL SYSTEM?

This is as close as I've gotten to reading Weera Ostapirat's (2004) article "Proto-Hlai Sound System and Lexicons" (sic) in 漢藏語研究龔煌城先生七秩壽慶論文集 Studies on Sino-Tibetan Languages: Papers in Honor of Professor Hwang-cherng Gong on His Seventieth Birthday. From what I can make out of a partial scan of p. 147, Ostapirat reconstructed this T-shaped 10-vowel system for Proto-Hlai:

(The high central vowel on p. 147 looks like *ɪ, but I wonder if the more common vowel *ɨ was intended. A lowered high central unrounded vowel ɪ is in only 4 languages in UPSID, whereas a nonlowered high central unrounded vowel ɨ is in 61 languages. I prefer to reconstruct more common segments unless extraordinary evidence compels me otherwise.)

Does any other language have a T-shaped vowel system? By 'T-shaped' I mean that it has a lot of high vowels but no nonhigh front or back vowels. Norquest's (2007: 331) 11-vowel system for Proto-Hlai (PHl) has mid vowels absent from Ostapirat's:

This system could also be described as T-shaped, though I normally don't think of Ts as having a thick horizontal stroke atop a stubby base.

Norquest's vowels have curious asymmetries:

- *ee and *aa are always long

- *o is always short and cannot occur in open syllables (since short vowels must be followed by a coda)

I wish I could read his dissertation to understand the reasoning behind these choices. This is not to say that I assume all vowels must be both short and long or that length must be distributed in a symmetrical manner: i.e., that both mid vowels should have the same lengths. Even languages with symmetrical vowel systems may have asymmetrical vowel frequencies: e.g., I think Japanese ee, oo, uu are more common than ii and aa is the least common long vowel of all. (The first three are in huge layers of Chinese loans, the fourth is mostly in recent loans and one class of native adjectives, and the fifth is almost entirely in recent loans*.)

I've never seen any inverted T-shaped vowel systems like

that are missing the top 'corners' of the vowel triangle (i, u) or the mid vowels beneath them (i, e).

Beekes' (1995:124) Proto-Indo-European vowel system is the shape of a tilted colon without any high or low vowels:

Does any attested language have such a system?

Even if one interprets his consonants (!) */i/ and */u/ as vowels, his system would still have a shape I've never seen in any other language:

(Beekes derived long *ii and *uu from *iH and *uH, so he did not reconstruct *ii and *uu as phonemes. Perhaps all long *e and *o also derived from *VH sequences.)

Edwin G. Pulleyblank (1965: 98) proposed a vertical system for Proto-Indo-European like the one he reconstructed for Old Chinese:

Such a system is reminiscent of the vertical system in some analyses of Kabardian: e.g., Matasović (2010: 9):

According to Matasović, short a and long aa also differ in "quality": aa is "low" whereas a is "central". Did he mean 'mid'? If so, is a between aa and ə in terms of height?

aa is also lower than a in Sanskrit, Vietnamese, and Cantonese. It seems there is a correlation between length and 'extremity': e.g., in English the higher vowels [ii uu] are longer than [ɪ ʊ]. Is there a language with a vowel system like

in which the highest and lowest vowels are shorter than vowels in the middle?

*Exceptions include the exclamation aa, (o)kaasan 'mother', and maamaa 'so-so'.

11.9.20.22:04: IS JIAMAO HLAI?

In "One Feathered Bear",I mentioned Norquest's (2007: 391) Proto-Hlai *tɕhɯɯ 'one'. 加茂 Jiamao, spoken by about 6% of Hlai, normally considered a Hlai dialect or language, has kɯɯ 'one' with a different initial. Jiamao in general is very aberrant compared to other Hlai varieties. Until I read Norquest's summary of his dissertation last night, it had never occurred to me that Jiamao could simply be non-Hlai - and non-Kra-Dai:

The language Jiamao is examined in detail, and it is argued that Jiamao is a non-Hlai language which has been in close contact with Hlai since the Pre-Hlai period. An examination of the correspondences between Jiamao and Hlai reveal at least two layers of Hlai loanwords in Jiamao, and evidence Jiamao was originally very different from Hlai structurally.

On page 14, he stated that Jiamao is "a non-Hlai language isolate". This idea goes back to Thurgood (1991b) which I haven't seen.

(How I wish I had spoken to Graham Thurgood about that when I was with him in the 90s! But back then I had only the faintest idea of what Hlai was. I didn't take a good look at Hlai until 2008 after Ostapirat (2000) got me interested in non-Tai Kra-Dai languages.)

Much of historical linguistics is about demonstrating relationships between languages. But how does one demonstrate a nonrelationship between languages? Google Books does not permit me to read Norquest's full answer, but from what I have seen:

- he interprets irregular correspondences between Jiamao and (other) Hlai languages as evidence for strata of borrowing.

- he also noted that Jiamao has non-Hlai basic vocabulary

Does this mean that Jiamao can or should be ignored in future Hlai studies? No, to the contrary,  as Norquest wrote (p. 398),

As Thurgood (1997: fn. 7) points out, Jiamao is of extreme value in the reconstruction of Proto-Hlai [...] It is vexing that the origin of the native Jiamao vocabulary is not more forthcoming, but future research (including non-linguistic data) may yet provide further clues into the elusive origin of this Southeast Asian language isolate.

I am always interested in loanwords, particularly in cases of massive borrowing like that between Chinese and its neighbors (e.g., Vietnamese, Korean, Japanese, Khitan, and of course, Tangut). Borrowing languages may preserve traits lost in the source languages: e.g., English chief retains an affricate [tʃ] that later became a fricative [ʃ] in French.

Why Hlai?

Although I am most interested in complex writing systems like Tangut, I still keep an eye on unwritten languages like Hlai because I hypothesize that they underwent sound changes similar to those in Chinese and Tangut. What happened in one language could have happened in others: e.g., intervocalic obstruent lenition occurred in both Hlai and Tangut:

Proto Hlai *-p- *-t- *-k- > Hlai w- l- Ø-

Pre-Tangut *-p- *-t- *-(t)s- *-(t)ʃ- *-k- > Tangut v- l- z- -ʒ- ɣ-

Vietnamese v- d- gi- g- also partly originated from intervocalic obstruent lenition.

Like Sagart (1999: 25-26), Norquest (2007: 313) proposed that a similar 'softening' also occurred in dialects of Northern Min Chinese.

I consider softening to be one kind of compression artifact - a remnant of an earlier, longer form. Phonetic features once 'loosely' distributed among a sequence of simple syllables have been partly retained with alterations in later, 'tighter' monosyllables.

11.9.19.23:48: ONE FEATHERED BEAR

In "Dog and Bug Are One", I wrote,

Schuessler (2007: 217) noted that 獨 was "a Han period 'southern Chu' dialect word for 'one'" and "may be related" to Fuzhou sioʔ < *dok 'one'.

I forgot to elaborate on "southern Chu". I recommend Wolfgang Behr's "Some Chŭ 楚 words in early Chinese literature" as an introduction to the Chu language.

Two phonetics in 'one'?

On p. 8 of Wolfgang's PDF is a Chu character

羽 *waʔ 'feather' atop 能 *nəŋ 'capable' = 'one'

I don't understand the function of 羽 'feather'. I would have expected 一 'one' as a semantic element. Could 羽 be a second phonetic representing a presyllable like *wə-? The structure of 'one' may be similar to that of Vietnamese double-phonetic characters like

𢁑 trái < blái 'fruit' < 巴 ba + 賴 lại

The phonetic element 能 *nəŋ was originally 'bear', but was used for an unrelated, difficult-to-draw homophone 'capable'.

A Tai tie?

能 *nəŋ sounds like Proto-Tai *hnɯŋ 'one'. Perhaps ?*wə-nəŋ and *hnɯŋ share a root like *nəŋ. PT *hn- (reflected in the spelling of Thai หนึ่ง <hnɯ1ŋ> 'one') may come from some earlier cluster of a voiceless consonant(al prefix?) plus *n-. Another possibility is that *hnɯŋ is the bare root and that the southern Chu word was something like ?*wə-hnəŋ.

Zhuang nwngh 'one', a descendant of PT *hnɯŋ, was written as the right side of 能 in the traditional sawndip script.

獨 *dok is obviously not related to PT *hnɨŋ but it does vaguely resemble Proto-Tai *ʔdiau 'single' and Proto-Kam-Sui *do 'one' which may or may not be cognate to each other. (PKS is a sister of PT. Both are descendants of Proto-Kam-Tai.)

The Kra-Dai family (Ostapirat 2000: 1)

Could 獨 *dok 'alone' be an early Tai borrowing into Chinese rather than a derivative of the Chinese root *tek? I doubt it because 獨 *dok 'alone' is in the Shijing predating Chinese-Tai contact and the phonetic match with Tai is weak. However, pre-Fuzhou *dok 'one' might be a borrowing from Proto-Kam-Sui *do or some cognate of it and the resemblance between pF *dok 'one' and Old Chinese 獨 *dok 'alone' may just be coincidental.

Proto-Kra-Dai 'one'

Can such a word be reconstructed? The *d-word(s?) shared by PT and PKS could be projected up to the PKT level but they do not resemble PT *hnɯŋ, Ostapirat's (2000: 245) Proto-Kra *tʂəm 'one' or Norquest's (2007: 391) Proto-Hlai *tɕhɯɯ 'one'. Are the last two words related? What's the Be word for 'one'?

11.9.18.23:12: DOG AND BUG ARE ONE

I left out one other possible member of the 'one' word family in Chinese from my last two entries:

獨 Old Chinese *dok 'alone'

now simplified as 独 < 犭 'dog' (why? 'lone wolf' comes to mind) + 虫 'bug', a reduction of the original phonetic 蜀 *dok 'silkworm-like caterpillar'.

This could be from an even earlier *N(ʌ)-tok sharing a root *t-k with words from my previous entries.

Why is the root vowel *o instead of *e?

Perhaps a labial presyllabic vowel conditioned rounding of the root vowel:

*No-tek > *No-tok > *Ntok > *Ndok > *dok > *dok?

Why is 獨 *dok emphatic unlike its phonetic 蜀 *dok?

Syllables with *o became emphatic (pharyngealized; indicated by underlining) in later Old Chinese unless preceded by a presyllabic high vowel. Compare:

蜀 *Cɯ-dok (perhaps *N(ʌ)-tok?) > *dok (high vowel presyllable > no emphasis)

獨 *N(ʌ)-tok > *dok (low vowel presyllable or no presyllabic vowel > emphasis)

燭 *Cɯ-tok > *tok 'torch' (high vowel presyllable > no emphasis)

with 火 'fire' on left

How was 獨 *dok phonetically distinct from 蜀 *dok?

I used to think 獨 *dok was *[dˁɔˁq] and 蜀 *dok was *[dok], but now I'm not so sure because the two words belong to the same Old Chinese rhyming class:

Would *[ɔˁ] and *[o] rhyme? Maybe. [ɔ] (sans pharyngealization) and [o] can interrhyme in Vietnamese poetry: e.g., mòn [mɔn] 'worn down' and chồn [con] 'tired' rhyme on p. 16 of the 1866 edition of 傳翹 Truyện Kiều.

Would *[q] and *[k] rhyme? Is there any language that allows them to rhyme?

Now I am inclined to reconstruct *dok as *[dˁoˁkˁ], though pharyngealized [kˁ] is an unusual sound. I only know of two languages with [kˁ]: Shilha (the only one listed with [kˁ] in UPSID) and Palestine Arabic. The Wikipedia entry for Shilha lists [q] instead of [kˁ]. I don't know of any language with a three-way phonemic distinction between /k/, /kˁ/, and /q/. Islam Youssef (2006) observed [k], [kˁ], and [q] in Cairene Arabic, but regarded the first two as allophones of /k/ contrasting with /q/.