15.11.21.23:59: WHY DID KOREANS BORROW LATE MIDDLE CHINESE GRADE II VELAR-FINAL SYLLABLES IN TWO DIFFERENT WAYS? (PART 1)

The stages [of Chinese presented here] are 'generic'; as I will demonstrate later, actual dialects could differ from this model.

Here's such a brief demonstration. What I reconstructed as Late Middle Chinese (LMC) *æ normally corresponds to Sino-Korean (SK) a. But that *æ also has other correspondences:

更 'change': MOC *kˀraŋ > LMC *kæŋ : SK kʌjŋ (not *kaŋ)

羹 'soup': MOC *kˀraŋ > LMC *kæŋ : SK kʌjŋ

庚 'seventh Heavenly Stem': MOC *kˀraŋ > LMC *kæŋ : SK kjəŋ < *keŋ (not *kaŋ or *kʌjŋ)

耕 'to plow': MOC *kˀreŋ > LMC *kæŋ : SK kjəŋ < *keŋ

客 'guest': MOC *kʰˀrak > LMC *kʰæk : SK kʌjk (no aspiration; borrowed before Korean developed a phonemic aspirated /kʰ/;  not *kak)

格 'go to': MOC *kˀrak > LMC *kæk : SK kjək < *kek (not *kak or *kʌjk)

隔 'obstruct': MOC *kˀrek > LMC *kæk : SK kjək < *kek

革 'hide, skin; change': MOC *kˀrək > LMC *kæk : SK hjək < *hek (irregular initial)

(The SK forms are premodern and in IPA to facilitate comparison. *-k forms added 11.22.23:34.)

I think ʌj and jə (< Old Korean *e) reflect two different aprpoaches to borrowing a northeastern LMC diphthong *ʌe corresponding to *æ before velars in other Middle Chinese dialects. ʌj matched the first vowel of *ʌe and approximated the second with a glide. Old Korean *e matched the second, more prominent half of *ʌe (which I could more precisely write as *ʌ̆e).

11.23.13:34: If you look carefully, you can see a pattern among the SK readings that I missed when I wrote this entry. I'll reveal that pattern next time.

15.11.20.23:59: HOW DID MEOW BECOME ME-NG-OW?

While looking through Thurgood's From Ancient Cham to Modern Dialects (1999) for examples of diphthongs resulting from vocalic splits*, I found two unusual-looking forms:

Western Cham maŋiau 'cat' from Proto-Chamic *miaw (p. 159)

Phan Rang Cham pimaw 'mushroom' from Proto-Chamic *bɔh maw (p. 158)

'Cat' has a -ŋ- that 'grew' in the middle of *miaw. Moreover, there is an extra -a- between m- and this -ŋ-. Why was this originally monosyllabic word stretched into two syllables when the general tendency was to compress? Examples from Thurgood (1999: 112):

Proto-Chamic *bara > Western Cham pra 'shoulder'

Proto-Chamic *bulan > Western Cham ea plan 'moon, month' (What does ea mean?)

Proto-Chamic *bulu > Western Cham plau 'body hair'

'Mushroom' disturbs me because i and *ɔ are very different vowels. I fear that Old Chinese presyllabic vowels might have undergone nonrecoverable changes similar to those that occurred in this word.

11.21.23:46: I am also disturbed by Phan Rang Cham mɨyau 'cat' (p. 159). Could this word be like a pre-Western Cham *mayiau, a stretched form whose intrusive *-y- nasalized to *-ɲ- under the influence of the preceding *m-? But why would  *-ɲ- back to -ŋ-?

11.22.0:19: Western Cham does have ɲ [Thurgood 1999: 274], though I do not know if word-medial -ɲi- is possible. Did *-ɲi- become -ŋi-?

15.11.19.23:59: CHINESE GRADE II, VERSION 2015.11.19

(More like version 2015.11.21, since I have revised this entry over the past two days.)

In my last entry, I reconstructed the Late Old Chinese reading of the Grade II word 講 'discuss, explain' (later 'speak') as *kəoŋʔ. I think I've reconstructed Grade II words with similar diphthongs before. In any case, here's how I think *-ˁr- (> Middle Chinese Grade II) and *-r- (> Middle Chinese Grade III) syllables developed between Middle Old Chinese (after emphasis had developed) and Late Middle Chinese:

The stages are 'generic'; as I will demonstrate later, actual dialects could differ from this model.

I wrote *-ˁ- before *-r- since I follow Baxter and Sagart in writing emphasis before the first consonant of a cluster: e.g., *pˁr-. But I think emphasis was a feature of all consonants in a cluster: e.g., /pˁr/ was phonetically [pˁrˁ].

Between Middle and Late Old Chinese, nonemphatic *-r- became a high central *ɨ that fused with nonfront high vowels, but emphatic *-ˁr- became lower-mid back *ʌ before lower vowels. (I decided to write the reflex of *-ˁr- as *-ʌ- because emphasis is associated with backing and because *-ˁrə did not become long *-əː.)

Perhaps vocalization predated vowel bending: e.g., *-re > *-ɨe > *-ɨie.

In Late Old Chinese, lower series vowels dissimilated:

*ʌa (achromatic-achromatic) > *ea (palatal-achromatic) : ae (!) in Baxter's Middle Chinese notation

*ʌə (achromatic-achromatic) > *ʌe > *ae (achromatic-palatal) : ea (!) in Baxter's Middle Chinese notation

*-ʌow (achromatic + labial + labial) > *-ʌew > *-aew (achromatic + palatal + labial) : aew in Baxter's Middle Chinese notation

In Early Middle Chinese, lower series diphthongs monophthongized

*ea > *æ

*ae > *ɛ

In Late Middle Chinese, *ɛ merged into *æ.

The first vowel in the achromatic-achromatic diphthong *ɨə dissimlated. Then the resulting fused into a front mid labial vowel that raised:

*-ɨə > -iə > *-ø > *-y

Go-on -o < *-ɨə, Kan-on -yo < *-iə, Sino-Korean -ŏ < *-ø, and Sino-Vietnamese -ư < *-y reflect these four stages. (However, those four types of Sinoxenic were borrowed from four different Middle Chinese dialects at four different times, so it is possible that they reflect different dialect developments: e.g., Sino-Korean -ŏ could be from a conservative *-ɨə or even *-ə and Sino-Vietnamese -ư, an unrounded vowel, may directly reflect an *-ɨ from *-ɨə.)

I am not happy with the diphthongs I reconstructed. I should examine the diphthongs of Mon-Khmer languages which have undergone vocalic splits and diphthongization to get a feel for which vowel sequences are plausible. (The only such Mon-Khmer language I am familiar with is Khmer which does not have ʌa, etc.)

11.21.22:42: I originally intended to include onset development in the table above but omitted it to focus on the vowels. Here are a few examples of how *r-clusters changed over time:

*ɨa became *ua after labials in syllables with zero or glottal codas: *Pɨa(H) > *Pua(H).

In late Early Middle Chinese, final mid vowels in diphthongs raised (and merged with the preceding vowel if it was identical):

*uo > *uu > *u

*ɨe > *ɨi

*ie > *ii > *i (no examples in this post because *ie was only in syllables that did not have medial *-r-)

Coronals developed retroflex allophones before rhotics:

*t(ʰ)r > *ʈ(ʰ)r > *ʈ(ʰ)

*dr > *ɖr > *ɖ

*nr > *ɳr > *ɳ

*(t)s(ʰ)r > *(t)ʂ(ʰ)r > *(t)ʂ(ʰ)

*(d)zr > *(d)ʐr > *(d)ʐ

Those allophones became phonemic after the rhotics were lost.

Final glottal stops conditioned glottalization in Early Middle Chinese which in turn led to a tone in Late Middle Chinese. (I think it might be better to translate 聲 as 'phonation' rather than as 'tone' for Early Middle Chinese.)

Labials weakened to dentilabial fricatives before *u: *pu > *fu.

Voiced obstruents became voiceless-obstruent-*ɦ-clusters in Late Middle Chinese (Pulleyblank 1984).

15.11.18.23:59: IS SANYA MANDARIN MANDARIN?

Today I found Thurgood et al.'s A Grammatical Sketch of Hainan Cham and was astonished by Appendix C on 三亞 Sanya Mandarin. An informant read a Mandarin newspaper out loud with surprising results:

- the codas included -k and -t unlike any Mandarin variety I have ever seen but like southern Chinese languages: e.g.,

國 kok²⁴ 'country' (cf. LOC *kwək, standard Mandarin guó)

雪 siet²⁴ 'snow' (cf. LOC *swɨat, standard Mandarin xuě)

but LOC *-p corresponds to *-t:

葉 zet²⁴ 'leaf' (cf. LOC *jɨap, standard Mandarin yè)

十 sit²⁴ 'ten' (cf. LOC *dʑɨəp, standard Mandarin shí)

- the codas included glottal stops and nasal-glottal stop clusters generally where they would be reconstructed in Old Chinese (!): e.g.,

九 kiuʔ⁴ ³ 'nine' (cf. LOC *kuʔ, standard Mandarin jiǔ)

講 kiaŋʔ⁴ ³ 'speak' (cf. LOC *kəoŋʔ, standard Mandarin jiǎng)

but see the exceptions below!

- "the numbers are all read in Cham except dates": e.g.,

六十年 as Cham naːnʔ³³ piu⁵⁵ tʰun³³ instead of Chinese *lok²⁴ sit²⁴ nien²¹

How much does this reading pronunciation correspond to the informant's spoken Mandarin? Is it possible that this reading pronunciation is derived from a conservative southern Chinese language?

15.11.19.21:31: There seem to be at least four strata in the reading pronunciations.

The first two strata (a conservative southern Chinese language and Cham) are listed above.

The third stratum looks like Mandarin and must be recent. It is characterized by an absence of final stops (and tonal differences I will explore later):

社會主義 se³³ huj³³ tsu²¹ zi³³ 'socialism' (cf. the LOC readings *dʑiaʔ, gwas, tɕuoʔ, and ŋɨajh - of course there was no LOC word 'socialism' - and standard Mandarin shèhuì zhǔyì)

but 主席 tsiuʔ⁴ ³ si³³ 'chairman' (cf. the LOC readings *tɕuoʔ and *ziak) preserves the final glottal stop in the morpheme 主 'master' - though the following morpheme lacks *-k!

得 tə²¹ 'get' (cf. LOC *tək and standard Mandarin dé)

說 s(i)o²¹ 'speak' (cf. LOC *ɕwɨat and standard Mandarin shuō)

The word 學習 sioʔ²⁴ sit²⁴ 'to study' is a combination of forms from different strata like 'chairman'. However, in 'chairman' the first syllable was more archaic, whereas in 'to study' the second syllable sit²⁴ is more archaic (cf. LOC *zɨəp and standard Mandarin xí without a stop coda). sioʔ²⁴ (< *xioʔ < LOC *gəuk) is like 合肥 Hefei Mandarin ɕyɐʔ with a glottal stop that is a trace of the *-k preserved in the oldest Chinese stratum. Hence I think sioʔ²⁴ is from a fourth stratum that is slightly more archaic than standard Mandarin xué which has no final stop.

15.11.17.21:10: IS 'WING' FROM 'BRANCH' IN CHINESE?

In my previous entry, I reconstructed Old Chinese

*Cɯ.ke > *ke 'branch' (spelled 支 or 枝 with 木 'tree'), limb' (spelled 肢 or 胑 with 肉 'flesh')

Could *C have been *s- if 'limb' is related to 翅 *sɯ.ke-s 'wing'?

Baxter and Sagart (2014: 140) reconstructed 翅 as Old Chinese *s-kʰe-s and *kʰe-s with an aspirated *kʰ- absent from their reconstruction of 'branch' as *ke.

Below I reconstruct a single 翅 *sɯ.ke-s 'wing' that underwent three different paths of reduction. I have included 'branch' and a possible cognate 咫 'foot (8 inches)' (< 'length of a branch'?) for comparison.

Early *s.k-reduction is part of the same wave of changes as *C.l-reductions 2-3 in this post. Similarly, late *s.k-reduction is part of the same wave of changes as *C.l-reductions 4-5.

If *sɯ- is part of a root 'branch', then *sɯ- was lost in 'limb, branch' prior to late *s.k-reduction whereas it was never dropped in the derived word 'wing'.

A wild possibility is that *sɯ- was lost in 'limb, branch' very early on - even before emphasis - but the resulting *ke did not become emphatic due to analogy with 'wing'. However, I would expect 'wing' to be remodeled after 'branch' rather than the other way around. In English, 'branch' is more common than 'wing'. Was the frequency the other way around in early Chinese?

15.11.16.23:54: WHY DOES ARABIC HAVE EMPHATICS IN LOANWORDS FROM LANGUAGES WITHOUT EMPHATICS?

Normally if a language with sound X borrows from a language without sound X, I wouldn't expect sound X to be in borrowings. So for instance Hindi-Urdu has voiced aspirates and English doesn't. Hence I wouldn't expect voiced aspirates in Hindi-Urdu loanwords from English. (If there are such loanwords with gh, etc., I'd like to know about them.)

However, Hindu-Urdu loanwords from English do have retroflex stops even though English doesn't: e.g., ḍākṭar from doctor.

The reason is that Hindi-Urdu lacks alveolar stops, and to Hindi-Urdu speakers, English alveolar stops are perceived as being closer to Hindi-Urdu retroflex stops than to Hindi-Urdu dental stops.

I have long known that Arabic has emphatics (in bold) in loanwords even from languages without emphatics: e.g.,

Latin strata > Greek strata > Aramaic ʕstˁrtˁ > Arabic sˁiraːtˁ 'way'

(11.17.19:59: I presume the Aramaic form had a vowel added before s to break up the initial cluster: str- > ʔVs.tˁV.r-. Did stˁ really simplify to sˁ in Arabic? Could the Arabic word be from Middle Persian <slt> /srat/ 'street'?)

French bicyclette > Moroccan Arabic bəqʃliːtˁa

Spanish falta > Moroccan Arabic faːltˁa 'error, offense'

French automobile > Moroccan Arabic tˁuːmubiːl

French déserter > Algerian Arabic zarˁtˁa

French exercice > Algerian Arabic garˁsˁiːsˁ (with French /gz/ simplified to /g/)

Italian gelati > Tunisian Arabic ʒiːlaːtˁ 'ice cream'

Turkish abla > Egyptian Arabic ʔabˁla 'older sister'

French tante > Egyptian Arabic tˁantˁ

Those examples are from Kossmann's chapter on borrowings in Owens (2013). I didn't see any explanation for those emphatics there, but I guess that they might have something to do with approximating foreign vowel qualities: e.g., it would make sense to borrow automobile as tˁuːmubiːl if /u/ had a lower allophone [ʊ] after emphatic /tˁ/ that was closer to foreign o than the higher allophone of /u/ after nonemphatic /t/.

Perhaps I am on the right track - at least for Moroccan Arabic (MA). Kenstowicz and Louriz wrote in their abstract:

MA has three vowel phonemes /a/ /i/ /u/ (as well as an epenthetic schwa). They take lowered and retracted allophones [ɑ], [e] and [o] respectively, when tautosyllabic with an emphatic consonant. The latter are redundant and predictable variant of the corresponding phonemes. This would lead one to predict that they should play no role in loanword adaptation. Also, since French lacks emphatic consonants, we expect that the above mentioned allophones should be absent completely from French loanwords in MA. However, consideration of French loanwords in MA shows that French /ɑ/, /e/ and /o/ can be identified with the MA allophones that appear in emphatic contexts in the native phonology.

I think this is the full article. I haven't had time to read it yet.

Note that the three vowels match the three 'lower series' vowels *a, *e, *o of my Old Chinese reconstruction which condition emphasis unless preceded by the 'higher series' vowels: e.g.,

雞 *ke > *kˁe 'chicken' (source of White Hmong qaib)

but *Cɯ.ke > *ke 'branch' (spelled 支 or 枝 with 木 'tree'), limb' (spelled 肢 or 胑 with 肉 'flesh')

An understanding of the apparent emphatic-nonemphatic mismatches in Arabic loanwords and their sources may help understand similar apparent mismatches in Old Chinese words and related forms in neighboring languages: e.g., between Old Chinese *kˁ- and White Hmong q- in 'chicken' (though to be a C > Hmong-Mien loan) and 'dog' (possibly a Hmong-Mien > Chinese loan?).

15.11.15.23:59: HOW DID HMONGIC AND MIENIC GET THEIR WORDS FOR 'IRON'? (PART 1)

The short answer is "from Chinese". Here's a longer answer. Thanks to Mark Alves for drawing my attention to this issue.

Baxter and Sagart‘s (2014: 160) reconstruction of the Old Chinese word 鐵 for 'iron' does not quite match the Proto-Hmongic and Proto-Mienic forms reconstructed by Ratliff (2010: 258):

I have rewritten Ratliff's hl- and hr- in IPA to facilitate comparison.

'Tone' is in quotation marks since Old Chinese did not have tones and it is not clear whether the other two proto-languages had them. Nonetheless these 'tonal' categories definitely have tonal reflexes in daughter languages.

The Proto-Hmongic form has a labial vowel *u where I would expect a palatal vowel.

Similarly, Proto-Hmongic has *u in *ʔjuw C 'small, young' borrowed from some reflex of Old Chinese 幼 *[ʔ](r)iw-s. I would have expected Proto-Hmongic *ʔjiw C, but there is no Proto-Hmongic rhyme *-iw in Ratliff's reconstruction.

Proto-Hmong-Mien *-k words developed Proto-Hmongic tone C unlike Proto-Hmong-Mien *-t and *-p words which developed Proto-Hmongic tone D (Ratliff 2010: 31). I suspect Proto-Hmong-Mien *-k became pre-Proto-Hmongic *-x which merged with *-h, the source of Proto-Hmongic tone C:

(11.16.1:19: This merger has a parallel in Chinese:

Unlike the pre-Proto-Hmongic merger, the Chinese merger involved clusters and perhaps a chain shift if Middle Old Chinese *-s really was [s]: *-ts, *-ps > *-s > *-h.)

Old Chinese *l̥ˁik 'iron' was borrowed before the merger as pre-Proto-Hmongic *l̥ix after Proto-Hmong-Mien *-ik and *-ek had become pre-Proto-Hmongic *-ɨx.

*-ix > *-iɣ > *-iɰʰ > *-iwʰ > Proto-Hmongic *-uw C

(11.16.1:26: Was Proto-Hmongic *-uw phonetically [ʊw]? Modern reflexes include [o], [ɔ], [ə], and [aw] as well as [u]. See Ratliff 2010: 135-136.)

Perhaps *-x (from an earlier *-k) generally became pre-Proto-Hmongic *-w which was then lost after certain vowels: e.g.,

*-ɨx > *-ɨwʰ > *-ɨ C (there was no *-ɨw in Proto-Hmongic).

(11.16.0:16: I am reminded of pre-Tangut *-k which shifted to *-w which was then lost after certain vowels: e.g.,

*-ak > *-aw > -a [there was no -aw in Tangut]

However, this secondary -w in Tangut was not associated with a particular tone unlike my proposed secondary *-w in Proto-Hmongic.)

The Proto-Mienic form has *r̥- instead of *l̥-. A rhotic also underlies Vietic forms for 'iron': e.g., Vietnamese sắt < *kr-. Did Proto-Mienic and Vietic borrow 'iron' from Chinese dialects in which *l̥- became *r̥-?

Next: Chronological issues.

11.16.1:10: The Vietic borrowing may reflect an archaic *kr- cluster from an even earlier *kʌ.l-:

*kʌ.lik > *kʌ.lˁik > *k.lˁik

> mainstream Chinese *l̥ˁik > *l̥ˁit > *l̥ˁeit > *tʰet

> dialect A *l̥ˁik > *l̥ˁeik (source of Proto-Tai *l̥ek 'iron' [Pittayaporn 2009: 333] and Proto-Palaungic *l̥ek 'iron' [Sidwell 2010])

> dialect B *k.rˁik > *r̥ˁik > *r̥ˁeik (source of Proto-Mienic *r̥ˁɛk)

> dialect C *k.rˁik > *k.rˁeik > *k.raik (source of Vietic forms; was -ik borrowed as a palatal stop *-c?)

(11.16.1:40: The shift *-ei > *-ai is in other southern reflexes of old emphatic syllables: e.g., 雞 *kˁe > *kei > *kai, the source of Proto-Tai *kaj B 'chicken' [the tone is unexplained and may reflect an *-h from an earlier *-s suffix in the source dialect]).

(11.16.13:32: Proto-Hmong-Mien *Kəi 'chicken' may also reflect a southern form like *kai.)

The Vietic borrowing may have displaced a native cognate of Proto-Katuic *taːʔ 'iron' (Sidwell 2005) if the Vieto-Katuic hypothesis (see Alves 2005) is correct.

Proto-Katuic *taːʔ 'iron' superficially resembles but is probably not cognate to Old Khmer <teka> /ɗɛːk/ 'iron' which Jenner compared to Siamese เหล็ก /lèk/ 'iron'. Could the Khmer form be from a Old Chinese dialect in which *k- dropped without conditioning the devoicing of the following liquid?

> dialect D *lˁik > *lˁeik > *dek

But why was Old Chinese *d- borrowed as an Old Khmer implosive /ɗ/ instead of /d/? Could dialect D have had implosives?

> dialect D *ʔlˁik > *ʔlˁeik > *ɗek?

Are there other instances of Old Chinese *lˁ corresponding to Khmer /ɗ/? I'd like to take a second look at Jenner and Pou's "Some Chinese Loanwords in Khmer" (1973).