08.6.21.23:59: HAINANESE DL- [D]

only appears before three finals (dli dlu dle) according to Xu and Yang. Why is it so rare?

Moreover, only one of the three examples has a known etymology: 你 Hainanese dlu < Middle Chinese *nɨʔ. dli was written as 些 MC *sjæ 'a few', so I presume their meanings are identical or at least similar, though they cannot be cognates. dle is a chorphan* without a cognate or translation equivalent in written Chinese languages. If dl- [d] < *nd- < *n- as implied by 你, what prevented dl- from developing before other finals?

*6.22.0:52: Defined by Bauer and Benedict (1997) as a

portmanteau word based on (Chinese) "character" + "orphan" to designate [a] type of morphosyllable in a Chinese dialect which does not have a standard Chinese character etymologically associated with it

I would also consider dli ?'a few' to be a chorphan since it has no etymological relationship with its apparent translation equivalent 些 MC *sjæ.

6.22.0:44: Xiamen and Chaozhou l- (also from *nd- < *n-) have a much wider distribution than Hainanese dl-: e.g.,

纳 Xiamen lap (Hainanese ?nap if it's homophonous with 钠 Hainanese nap 'sodium')

钠 Md na 'sodium' (cf. the symbol Na < Lat natrium) is a recently coined sinograph, so its reading is presumably by analogy with 纳

內 Xiamen lue (lit.), lai (col.), Chaozhou lai (Hainanese ?noe)

男 and 南 Xiamen and Chaozhou lam (南 is Hainanese nam; presumably 男 is also nam in Hainanese)

难 Xiamen lan, Chaozhou laŋ (Hainanese nan)

念 Xiamen liam, Chaozhou liəm (Hainanese niam)

Hainanese has b- < *mb- < *m- before 28 finals and gz- [g] < *ŋg- < *ŋ- after 10 finals. I would have expected all denasalized initials to have roughly the same distribution.

or 内爆破音 'inner explosive break sounds' are implosives: [ɓ ɗ]. Although such sounds are in Chinese-like languages (e.g., Hlai and Vietnamese also have [ɓ ɗ]), they aren't in any other Chinese language that I know of.

徐基 军 Xu Jijun and 杨文豹 Yang Wenbao romanized these sounds and Hainanese [g]* as bz dz gz to distinguish them from b d g representing the regular plosives (爆破音 'explosive break sounds') [b t k]. The use of voiced symbols for [t k] was carried over from standard Mandarin Pinyin. I think the use of z (which represents [ts] in standard Mandarin Pinyin) is misleading: bz dz gz are not [bz dz gz] or [pts tts kts]. I would recommend bb dd gg for [ɓ ɗ g]. There would be no ambiguous syllable divisions because Xu and Yang romanized final [p t k] as p t k. Hence a (hypothetical?**) word like 基地 giddi (Xu and Yang's gidzi) 'base' could only be gi-ddi rather than gid-di. Another possible romanization of [g] could be gh to distinguish it from the implosives bb [ɓ] and dd [ɗ], though gh implies [ɣ] or [gɦ].

Xu and Yang romanize Hainanese [d] as dl, but this could be changed to dh to match gh. (However -h- once again implies a fricative or voiced aspirate.)

Since dl originates from *n, a historically based spelling could be dn. nd would not be possible since CVndV could be interpreted as either CVn-dV or CV-ndV. A velar counterpart to dn would be the rather awkward-looking gng [g] < *ŋg < *ŋ.

6.21.0:43: There is no need to romanize [b] as bh or bm (reflecting how it orignated from *mb < *m) because there is no need to distinguish it from an initial romanized as b-.

*Xu and Yang regard [g] (their gz) as an implosive, but it sounds like a regular [g] to me. I presume Li Fang-kuei also heard a regular [g], as he only reported two implosives in Hainaese: [ɓ ɗ]. You can hear gz at 1:28 in their online video. The Hlai languages described by 苑中树 Yuan Zhongshu (1994) have [g] rather than implosive [ɠ] as the velar counterpart to [ɓ ɗ].

[g] is historically as well as phonetically unlike [ɓ ɗ]: the former is from an earlier nasal (as in other southern Min languages) whereas the latter are from earlier stops, as in Vietnamese:

疑 Hainanese gzi [gi]: Middle Chinese *ŋɨ

cf. other southern Min languages: Xiamen gi, but Chaozhou ŋi

边 Hainanese bzi [ɓi]: Middle Chinese *pen

cf. Vietnamese biên [ɓiən]

部 Hainanese bzu [ɓu]: Middle Chinese *boh

cf. Vietnamese bộ [ɓo]

担 Hainanese dza [ɗa] : Middle Chinese *tanh

cf. Vietnamese đán [ɗaan]

地 Hainanese dzi [ɗi]: Middle Chinese *dih

cf. Vietnamese địa [ɗiə]

**giddi is my hypothetical Hainanization of standard Mandarin jidi 'base'. I don't know if such a word actually exists in Hainanese.

6.21.0:29: I don't understand why there is no Hainanese syllable gza. I would expect syllables like 雅 Middle Chinese *ŋæʔ to correspond to Hainanese gza. (Cf. how 麻 MC *mæ corresponds to Hainanese ba.) I presume that 雅 is native Hainanese gze (like 牙 gze : MC *ŋæ) without a borrowed counterpart gza since MC *æ corresponds to both Hainanese -e (in native morphemes?) and -a (in borrowings?).

08.6.19.23:59: KAAN 'STEEL' BE A LOANWORD?

Another, somewhat less puzzling Chinese loanword in Yuan (1994: 48) is Hlai (variety unspecified) kaan¹(koŋ¹) 'steel'.

I don't understand the significance of the parentheses. I'd expect them to surround a nonloan morpheme in a Hlai-Chinese hybrid compound, but they surround koŋ¹ which looks more like Hainanese ?ko* ( < 鋼 Middle Chinese *kaŋ) 'steel' than kaan¹ does. Maybe Yuan was indicating that kaan¹ is an older loan and koŋ¹ is a newer loan. (These words were listed in his section on the oldest layer of Chinese loans in Hlai.)

The final -n of kaan¹ is strange. I would expect -ŋ instead, as -aaŋ is a permissible Hlai rhyme. Could -n be an attempt to reflect early Hainanese nasalization (i.e., a form like *kãa) rather than a segmental -ŋ?

At least the first tone in both kaan¹ and koŋ¹ is what I would expect from 'steel' which is a first tone word in Chinese.

*Implied by the Pinyin-style romanization go in 徐基军 Xu Jijun and 杨文豹 Yang Wenbao's 海南话拼音方案 Hainainese Pinyin Plan.

6.20.00:24: I'll take a closer look at this proposal next time.

08.6.18.17:25: KAN 'SILVER' BE A LOANWORD?

Yuan Zhongshu listed Hlai (variety unspecified) kan¹ 'silver' as a Chinese loanword in Liyu yufa gangyao (1994: 48). This has puzzled me for almost two weeks because I've never seen any Chinese word for 'silver' that looks like kan with a 'first tone'* (implying a *voiceless initial).

I don't know what the Hainanese word for 'silver' is, but I would assume that it's similar to words with 'second tones' (implying *voiced inititials) in other Southern Min dialects: e.g.,

Xiamen gun²

Chaozhou ŋɯŋ² (in Hanyu fangyin zihui notation), ŋɤŋ² (in Lin and Chen 1996, #1278)

Shantou ŋɤŋ²

Jieyang ŋeŋ²

Haifeng and Dianbai ŋin²

Haikang ŋieŋ²

Longtu nun² (sic!)

Older Chinese forms of 'silver' also don't sound like kan:

Middle Chinese *ŋɨin

Old Chinese *rŋən < ?*t-ŋər (cf. Written Tibetan dngul 'silver'

I know almost nothing about Hainanese, so I can only guess that the Hainanese word for 'silver' was something like *gan at the time of borrowing** if Hlai kan¹ really is a loanword. g- is a common Southern Min reflex of *ŋ-. The early Hlai may have added a prefix to this word that devoiced the initial, resulting in a first rather than a second tone after tonogenesis:

*C-gan > kan¹

Although some modern Hlai languages (Baoding and Tongshi) have a g-, I wouldn't expect 'silver' to be gan in those languages, since modern g- seems to be from earlier *ɣ- (see Ostapirat 2000: 60). I suspect a chain shift occurred:

*ɣ- > *g- > k-

It's also possible that

- the borrowing of 'silver' predated tonogenesis

- early Hainanese tone 2 sounded like early Hlai tone 1

Early Hainanese *gan² could have been borrowed as early Hlai *gan¹, which became kan¹ after initial devoicing. Another example of this tonal correspondence is 'money'

Middle Chinese *dzien > tone 2 (e.g., Xiamen tsĩ2, Chaozhou tshĩ2)

6.20.00:31: Xu and Yang list ji (phonetically [tsi] or [tɕi]?) for 钱 'money'

Hlai (variety unspecified) tsiin¹

However, not all Chinese tone 2 correspond to Hlai tone 1, and vice versa. Different strata of borrowing have different tonal correspondences which I'll investigate later.

Next: Another mystery metal.

*Yuan's comparisons on pp. 23-24 indicate that his Hlai tone 1 corresponds to tone category A1 (implying a *voiceless initial) in other Kam-Tai languages.

**6.20.00:27: 徐基军 Xu Jijun and 杨文豹 Yang Wenbao's 海南话拼音方案 Hainainese Pinyin Plan lists the reading ngin for 银 'silver'. This could be a literary loanword that displaced my hypothetical native *gan (if no such word exists in current Hainanese). If there are no instances of Hainanese -an corresponding to Old Chinese *-ən, then I can't see any way to link kan¹ to Chinese.

Since Yuan Zhongshu's Liyu yufa gangyao (1994) is the only book on Hlai I have, I'm not sure what to make of some oddities in it: e.g.,

- pi- is listed as an initial in the Baoding, Xifang, Baisha, Tongshi, and Jiamao varieties of Hlai (p. 1). Is pi- supposed to be a palatalized p-? No initials like ki-, etc. are listed. It seems unlikely that [pʲ] would be the only palatalized initial.

- pi- and ʔb- are listed together in the same row, unlike all other initials. Is this significant?

- So far, I have found three initials in the examples that do not appear in the table of initials:

tɬ- (in Jiamao; p. 166)

ʔdr- (in Baisha; p. 168)

pl- (in an unspecified Hlai language; pp. 188, 192-195, 198, 200-202, 204-206, 209-211, 214, 217, 219)

kb- in kbom (a conjunction) on p. 188 is a typo for kh- because the word appears elsewhere as khom. But how many other 'missing' initials are typos?

pl- seems to be a genuine missing initial. Since it appears numerous times in the first sample text in the words plei 'swim', pleɯ 'hear', ploŋ 'house' and ploŋ 'husband' (the latter two with different tones), I don't understand why it's not in the initial list.

Or is it? I just realized that pi- is probably a typo for pl-. Could pl- be the sole surviving Cl-cluster in some Hlai language(s)? I wonder if Jiamao and Baisha have other liquid clusters besides tɬ- and ʔdr-.

While looking up 目 'eye' in the Mandarin dialects in Hanyu fangyin zihui, I found a very strange set of readings unexpectedly ending in nasal codas:

A zero-nasal correspondence is not entirely unusual in Chinese: e.g., 耳 'ear' has both nonnasal and nasal versions:

nonnasal: Beijing er, Cantonese ji < ult. Old Chinese *nəʔ

nasal: Amoy dzĩ, Chaozhou hĩ < ult. Old Chinese *nəŋʔ

However, it is odd that Hankou and Chengdu have final nasals only in mo-syllables. Did m- (but not n-, ɲ-, or ŋ-) somehow independently trigger nasalization of o (but not other vowels) in two dialects that are distant from each other?

*mo > *mõ > *mõŋ > moŋ

That seems extremely unlikely.

Assuming that -ŋ is a retention lost in all other Chinese languages is also unlikely. It's possible that *-ŋ could be lost before *-ʔ and *-h and alternations of *-ŋ with *-k are not unknown. But how could *mot become moŋ?

The data in Hanyu fangyin zihui is handwritten, so I initially thought ŋ could be an error for u, but the two letters don't look alike (unless one elongates the bottom right part of u), and ŋ appears several times.

I can't explain these moŋ at all.

6.17.00:12: Added the Old Chinese column. I know of no premodern textual support for a nasal ending in 'eye', etc. External as well as Chinese-internal evidence points toward a *-k in 'eye'.

慕 and 幕: mɔ̃

墓: bɔŋ

酵 Amoy kã̃ : MC *kæwh < OC *krus

好 Amoy hõ, hɔ̃ : MC *xawʔ, *xawh < OC *xuʔ(s)

高 Amoy kuãi : MC *kaw < OC *kaw (are these words even cognate?)

賀 Amoy hɔ̃ : MC *ɣah < OC *gajs

他 Amoy thã : MC *tha < OC *hlaj

舀 Amoy ĩu : MC *jiewʔ < OC *lawʔ

which are not reconstructible with a nasal.

08.6.15.23:55: A TANGRAPHIC MISCELLANY

di R11 2.10 dzɑ R17 1.17

is a classified lexicon of Tangut.

Thanks to Mahaadaatṛ, I can finally see Sofronov and Terent'ev-Katanskij's 2002 edition which includes a Russian commentary and translation. Richard Cook's Tangut site has a scan of a page of Li Fanwen's 1997 dictionary which contains a sample page from the Miscellany at the bottom right. (The other three dictionaries in the scan are the Tangraphic Sea [top left], Homophones [top right], and the Pearl [bottom left].)

The section of the Miscellany that interests me most are the sections on names:

mi R11 2.10 R mɨʳ R92 1.86 mə R28 2.25

'Tangut people's surnames' (entries 533-777)

dzwio R53 2.44 miee R40 2.35

'[Tangut] people's [personal] names' (entries 778-825)

Here is a sampling of entries:

Tangut surnames

534. TT3343 ŋwei R8 2.7 TT1493 mi R11 1.11

ŋweimi was the imperial surname

incredibly, mi was analyzed in Tangraphic Sea as mi 'not' + 'holy'!

these tangraphs are pure surname tangraphs without any other meanings

535. TT2513 guu R5 2.5 ?TT3496 zwị R70 1.67

the second tangraph is similar to TT3496 but seems to have TT3087 'waist' on the left

TT3496 zwị means 'nephew' but may be a phonetic symbol here; TT2513 is only a surname tangraph, so guuzwị might mean 'Guu's nephew'

536. TT3912 ŋwəə R32 1.31 TT1789 dzõ R56 1.54

TT3912 can also mean 'incantation' and TT1789 is a pure surname tangraph: 'the incantation Dzo' (as opposed to other Dzo clans?)

Tangut personal names

779. TT3536 miə R31 2.28 TT5745 niaa R21 2.18 TT3672 bɛɛ R39 1.38

miəniaa is 'Tangut' and bɛɛ is 'amusement; happy' (Li Fanwen 1997), 'joy' (Terent'ev-Katanskij's translation); 'joy of the Tangut'?

780. TT3674 bə R28 2.25 TT2546 du R4 2.4 TT2761 ɣwɪ R9 1.9

'offspring of domestic animals'/(surname) + 'pagoda' + 'power': is this a full name (Be Dughwi = Pagoda-Power Be?) or a single name Bedughwi = 'power of Be's pagoda'?

781. TT3637 thɑ R17 2.14 TT3620 xwæ R18 1.18 TT3672 bɛɛ R39 1.38

'rely, depend' (LFW 1997), 'support' (n.; Terent'ev-Katanskij) + 'flower' (Chn. loan) + 'amusement; happy; joy'

is this a full name (Flower-Joy Tha), a phonetic transcription of a single name Thaxwabee, or 'Joy of the Support Flower'?

There are many names which defy analysis because they consist of pure onomagraphs devised solely to write names: e.g.,

the surnames Mbeiwon, Chye'on, Naizhye, Thawa, Toto (written with the same tangraph twice)

the names Xwashyendin, Khimupha, Keuzhi

Some surnames have prefixes with unknown meanings:

TT3378 si- (written as 'person' + 'woman') : Simi, Simbyu, Singyu, etc.

TT3836 lhye-: Lhyendin, Lhyembin (mbin = 'gold'), Lhyenga, etc.

TT5384 o-: Ozhye, Osan, Ose, etc.

The forms in the previous paragraphs were all in a simplified version of Sofronov's reconstructions. It's time-consuming to identify each tangraph and convert its reading into my system because Terent'ev and Sofronov's book has no Sofronov numbers (basically equivalent to the Tangut telecode used here).

I hope to get Li Fanwen and Nakajima Motoki's edition of the Miscellany which presumably has the AA number for each graph.