Thanks to Andrew West for identifying the language used in this stop sign that was puzzling me. He found the probable source of that uncredited photo.

I wonder if the sign is correct. Looking at this article, I don't see any references to

- nh in syllable or word-final position (it does exist in word-medial position; it is a cluster [n] + [h], not [ɲ] as in Portuguese or Vietnamese)

- a grave accent over n (or any consonant); the article gives the impression that a grave accent can only be atop a vowel followed by a colon. The grave accent represents a falling tone and the colon represents vowel length.*

However, absence of evidence is not evidence of absence.

*I presume Mohawk readers use context to distinguish between a word ending in a colon and a word without a colon followed by a colon. If the colon never appears in word-final position, ambiguity would not be an issue. URDREH AND SWEDISH: THE COLLECTIVE CONNECTION

Back in June, I proposed that the Urdreh language spoken on Robinson's fictional world of Hadanus distinguished between count and collective plurals: e.g.,

Urdleh (three or more Urdreh; count plural)

Urdreh (a large, unspecified number of Urdreh; the Urdreh people as a whole; collective plural)

On Thursday night, I learned that Swedish had a similar distinction for a "few nouns" according to Holmes and Hinchliffe (2008: 44): e.g.,

Gloss Singular Count plural Collective plural
pea ärta ärtor ärter
mosquito mygga myggor mygg
police polis poliser polis
man man män man/mannar


två ärtor 'two peas' (count plural)

ärter med fläsk 'peas with pork' (collective plural)

tre myggor 'three mosquitoes'

mycket mygg 'a lot of mosquitoes'

This distinction did not exist in Proto-Indo-European and it's not in German, Dutch, or English. Where did it come from?

The two plurals of man 'man' are reminiscent of English men and German männer 'men'.

mygg 'mosquitoes' (collective) has a zero ending whereas its singular ends in -a: mygga. This singular -a ~ plural zero alternation reminds me of Russian feminine -a nouns:

kniga 'book' : knig 'of books'

But note that this alternation has nothing to do with the Swedish one and that knig is a genitive plural, not a collective plural. A SHEEPISH SERIES (PART 2)

Here are several solutions to the question I posed in part 1. The last solution is my extrapolation based on long-time reader David Boxenhorn's suggestion that the common denominator of GSR 732 might've been *ɣ.

Example Initial type Middle Chinese Old Chinese reconstruction
Karlgren Sagart 1999 Baxter-Sagart Boxenhorn/AMR
1 *j- *zj- *l- *ɢ- *ɣ-
2 *z- *dzj- *s-l- *sɢ- ?*sɣ-
3a *k- *kj- *k-l- *Cə-q- ?*k-ɣ-
3b *kh- *khj- *k-hl- *Cə-qh- ?*s-k-ɣ-

Some other information:

Possibly related words in Tibeto-Burman language have j-.

Possible Chinese borrowings into Thai have l-.

Karlgren's reconstruction doesn't account for 3a and 3b or the possible Thai loanwords.

Karlgren derived 1 and 2 from a chain shift: *dzj- > *zj- > *j-.

The j-words in TB could be borrowings postdating the *zj- > *j- shift, but I doubt it.

Sagart's 1999 reconstruction accounts for all four initial types and the possible Thai loanwords. It too requires the TB j-words to be viewed as late loanwords, unless OC *l- and TB *j- are from a Proto-Sino-Tibetan *lj- (Schuessler 2007: 559).

More recently, Baxter and Sagart have proposed that GSR 732 is a uvular series. They reconstruct both emphatic and nonemphatic uvulars (does any language have such a distinction for uvulars) in OC and reconstruct nonemphatic uvulars for GSR 732. I presume that nonemphatic *ɢ- becomes a uvular fricative *ʁ- and then a velar fricative *ɣ- on the way to becoming Middle Chinese (MC) *j-.

I don't know whether B&S think that the TB j-forms are loans from Chinese postdating that shift or inherited from Proto-Sino-Tibetan *ɢ-forms.

Unknown prefixes conditioned *q(h)- to become velar.

If B&S are correct, the Thai l-forms are unrelated since GSR 732 never had liquid consonants in their reconstruction.

David Boxenhorn's *ɣ- is identical to the intermediate stage I posited for B&S' hypothesis. It would be the voiced counterpart of OC *x-. (But OC *s- has no voiced counterpart. Are there any languages with the fricatives /s x ɣ/ without /z/?).

*ɣ- could have independently become *j- in MC and TB.

*s- assimilated to *ɣ- by becoming voiced and then became a vowel (or velar glide) in MC:

*sɣ- > *zɣ- > *zɨ- (or *zɰ)

The group name prefix*k- (Sagart 1999: 107) did not voice before *ɣ-.

I reconstructed *s-k- for 3b because I suspect at least some Chinese aspirates are secondary. In this particular case, I wonder if the ethnonym 羌 (a drawing of a sheep with legs; now Mandarin Qiang) meant 'people of the sheep' and was derived from OC *ɣaŋ 'sheep', the group name prefix *k-, and another prefix *s- with an unknown function.

David's *ɣ- does not match Thai l-, though one could try to come up with a way to link the two:

OC *ɣ- > *ʁ- > *r- > *l- > borrowed into the ancestor of Thai as *l-

OC *skɣaŋ 'Qiang' resembles Written Tibetan skyong-ba 'to guard, keep, tend' even more closely than Schuessler's OC *kh(i)aŋ, even though Schuessler (2007: 426) suggested a link between the two words. Schuessler proposed that 羌 'Qiang' originally meant 'herders'. A SHEEPISH SERIES (PART 1)

The Chinese character 羊 'sheep' is phonetic in characters with three or four types of Middle Chinese (MC) initials:

1. MC *j-: e.g., 羊 MC *jɨaŋ 'sheep' itself and its homophones 佯洋痒, 養癢痒 MC *jɨaŋʔ, 恙羕漾樣 MC *jɨaŋh, etc.

2. MC *z-: e.g., 痒 MC *zɨaŋ (also *jɨaŋ above) and its homophones 庠祥翔詳 (which lack *j-readings)

3a. MC *k-: 姜 MC *kɨaŋ (no other examples)

3b. MC *kh-: 羌 MC *khɨaŋ (no other examples)

The graphs in this series (GSR 732) must have been pronounced similarly in Old Chinese (OC). What was their phonetic common denominator (i.e., their common consonant) in OC?

8.6.1:39: A hint?: Schuessler (2009: 83) lists Thai เลียงผา liaŋphaa '[mountain] goat' and เลี้ยง liaŋ 'nourish' as possible loanwords from Chinese 羊 'sheep' and 養 'nourish'. AGNES' WATER

Yesterday I couldn't figure out why amniotic fluid was translated in Chinese, Japanese, and Korean (but not Vietnamese*) as 羊水 'sheep water' until I looked up amnion at merriam-webster.com:

Etymology: New Latin, from Greek, caul [a native English word for amnion?] from amnos** lamb — more at yean [the English cognate of amnion; the ye- is a prefix, so the -an is all that matches]

But that begs the question: what does an amnion have to do with lambs?

*The Vietnamese term is nước ối, literally 'water amnion' = 'amniotic water'.

**Greek amnos is cognate to Latin agnus 'lamb' which in fact is not cognate to the name Agnes. Agnes is related to the hagio- of hagiography.

Why does -mn- in Greek amnos correspond to -gn- in Latin agnus? Greek -m- in -mn- was once a *-b- that assimilated to the following nasal. That *-b- in turn is from a Proto-Indo-European labiovelar *-gʷ- that became bilabial in Greek but velar (g) in Latin.

8.5.23:36: I have no idea why Germans and Dutch call amniotic fluid 'fruit water': Ger Fruchtwasser and Du vruchtwater. SHEEP WATER

Yesterday I tried to figure out the meanings of several Swedish compound words on Sarah's site:



fostervattnet (< fostervatten + definite article -et)


Two of them correspond to East Asian* terms that have Chinese components literally meaning

'sheep water'

'womb plate'

What do these terms mean?

*These terms are used in Chinese, Japanese, and Korean. Vietnamese has a different kind of 'water' term formed from native words, but I won't reveal it here because it'll give away the answer. EARLY KOREAN PENINSULAR NAMES: SPLIT FIRE?

This morning I came up with a far-fetched explanation for the Koryo Dynasty name 陽岳 'yang/sun high mountain' ('yang' as in yin and yang) for Paekche 分嵯, etc. If 分 ~ 夫 were transcriptions for Paekche *pur, and if the memory of Paekche *pur persisted until Koryo times, then perhaps Paekche *pur reminded someone of Middle Korean pɯr 'fire', and was very loosely translated as 陽 'sun', just as *tsai 'ridge' only vaguely corresponds to 岳 'high mountain'. EARLY KOREAN PENINSULAR NAMES: SPLIT RIDGE

Years ago, I had been planning to write a book on the earliest written evidence of languages (plural!) in Korea. I did some preliminary research and presented the results at a conference in Belgium in 2000 and have returned to the topic from time to time, but I'll probably never write the book. Posts like this one are all I can do now.

Korea was once divided into multiple states. Each is often assumed to have its own language. Records are very limited. We have no complete sentences - not even phrases or single words. We only have names which may have later replacements possibly hinting at what they mean.

For example, in the early peninsular kingdom of Paekche there was a place (Ryu Ryol #319) whose name was written as

分嵯 'split craggy' (嵯 is half of 嵯峨 'craggy')

分沙 'split sand'

夫沙 'man sand'

波知 'wave know'

When Paekche was conquered by Shilla, this place was renamed

分嶺 'split ridge'

Under the Koryo Dynasty it was renamed again as

樂安 'joyful peace'

陽岳 'yang/sun high mountain' ('yang' as in yin and yang)

Today, 樂安 is pronounced as 낙안 Nagan. Here's an overview of Nagan's history in English.

Although the pronunciation of its original Paekche name is unknown, the multiple spellings give us clues. When pronounced in Late Old Chinese, they mostly sound roughly similar, though the fourth barely matches:

分嵯 LOC *pun dzai

分沙 LOC *pun ʂai

夫沙 LOC *pua (later > *puo > *puə) ʂai

波知 LOC *pai ʈie

This indicates that these spellings are probably phonetic and that their meanings ('split craggy', etc.) are irrelevant.

Nagan's Unified Shilla period name 分嶺 was probably pronounced *punryOng - close to modern Korean 분령 pullyOng. (The consonant cluster *-nr- became -ll- in modern Korean.) 分 was carried over from two of the Paekche spellings and is presumably phonetic. 嶺 *ryOng 'ridge' is presumably semantic and a translation of a Paekche word written as 嵯 ~ 沙 (and 知?). Koryo 岳 'high mountain' could be a re-translation of 嶺 'ridge'. (樂安 'joyful peace' and 陽 'yang' seem to have nothing to do with the other names.)

In Korean, 'ridge' is 재 chE from Middle Korean tsai, which resembles the LOC pronunciations of 嵯 (*dzai) and 沙 (*ʂai), the graphs used to write what might be the Paekche word for 'ridge'. I conclude that Korean and Paekche have related words for 'ridge'. But was the Paekche word *dzai, *ʂai, or something else? I think it was *tsai as in Middle Korean.

However, why would *tsai be written with characters pronounced with *dz- and *ʂ- in Chinese?

First, I don't know of any Chinese character pronounced *tsai in LOC. No such character exists in Schuessler's 2009 LOC dictionary. Therefore Paekche scribes settled for a character with a *ts-like initial: e.g., 嵯 *dzai and 沙 *ʂai.

Second, there is no uncontroversial evidence for the consonants *dz and in Korean peninsular languages. Hence to Paekche ears - and perhaps also in Sino-Paekche pronunciation (i.e., Chinese with a Paekche accent), 嵯 LOC *dzai and 沙 LOC *ʂai may have sounded like *tsai and *sai.

Third, it is difficult to distinguish between the consonant sequences -nts- and -ns-: e.g., English prints and prince. Thus the Paekche might have thought that 嵯 Sino-Paekche *tsai and 沙 Sino-Paekche *sai sounded alike after the *-n of 分 Sino-Paekche *pun. The spelling alternation 嵯 ~ 沙 is therefore not evidence for the lack of a *ts : *s distinction in Paekche. If such a distinction were nonexistent, I would expect to find such alternations in initial position as well, but I don't.

Other aspects of the Paekche spellings are baffling:

1. Why does 分 LOC *pun alternate with 夫 LOC *pua and 波 LOC *pa? Perhaps these are attempts to write a Paekche *pur with a final *-r absent from LOC. 夫 LOC *pua (later > *puo > *puə) is like British English [puə] for poor. But 波 LOC *pa is unlike *pur except for a shared initial. Moreover, *ts and *s are easier to distinguish after *-r than after *-n. Compare the English nonsense word combinations ports' eye and pour sigh. The Paekche name may have been *purtsai, which is close to ports' eye.

2. 知 LOC *ʈie doesn't match the *tsai-archetype of 嵯 LOC *dzai and 沙 LOC *ʂai.

The spelling 波知 is the odd man out. Perhaps it stood for an unrelated Paekche name like *pate which happens to have the same initial *p- as *puntsai or *purtsai. THURGOOD'S SINO-TIBETAN CLASSIFICATION (PART 1)

Last night I mentioned an article on Phan Rang Cham which is just one out of many at Graham Thurgood's site. (My name turns up in this article on the origin of tones in Vietnamese. I remember talking to Graham about that, but I didn't know I was also mentioned in these two papers until tonight. The second of the three papers is a revision of the first; the third is different.)

Graham has added a new article since I was last there on Sino-Tibetan classification. (Oh no! Not another scheme! Ha, ha.) I haven't had time to really look at it yet, but I noticed that his classification of Chinese languages* doesn't include Jin (see "Hidden Jin-ius") as a branch or even as an example of a variety of Mandarin.

Graham's classification is based on Jerry Norman's (1988) classification with two major exceptions.

First, Graham doesn't use Norman's four-way classification of Mandarin dialects (northern, northwestern, southwestern, and eastern). His eight-way classification excludes Jin and matches this Wikipedia map:

东北话 Northeastern Mandarin

中原官话 Zhongyuan (Central Plain) Mandarin

北京话 Beijing Mandarin

兰银官话 Lan-Yin Mandarin

冀鲁官话 Ji-Lu Mandarin

西南官话 Southwestern Mandarin

胶辽方言 Jiao-Liao Mandarin

江淮官话 Jiang-Huai Mandarin

Names of the type X-Y contain abbreviations for two place names: e.g., Lan-Yin is short for Lanzhou and Yinchuan. Some short forms stand for unrelated longer names: e.g., Ji-Lu combines the old shorter names of Hebei and Shandong, not two places named Ji ... and Lu ...

For more on the Mandarin varieties without links above, see the Wikipedia article on Mandarin dialects.

Norman (1988) doesn't even have Jin in his index, so it's not surprising Graham doesn't mention it either. This just goes to show how hidden Jin is.in spite of its 45 million speakers - more than Polish! Norman would consider Jin to be a variety of northwestern Mandarin rather than a separate branch of Chinese.

Second, Graham groups Hakka under Gan in the central group rather than alongside (not under) Yue (e.g., Cantonese) and Min (e.g., Taiwanese) in the southern group.

In any case, what I'd like to see is a classification of Chinese languages - and Sino-Tibetan languages in general - based on shared innovations in each branch. I'll expand on this next time.

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