(Recommended background reading: glottalic theory, ejective consonant, implosive consonant, glottalic consonant.)

Last Saturday's post on ejectives in Ehret's Proto-Afroasiatic and Proto-Semitic consonant reconstructions got me thinking about ejectives in Proto-Indo-European (PIE).

Traditional reconstructions of PIE have three series of stops. Let's call these three series I, II, and III.

I. voiceless: p

II. voiced: b

III. voiced aspirate:

Here's my take on the reinterpretation of these series according to the glottalic theory.

The earliest stage of PIE had three series like Georgian or Ehret's Proto-Afroasiatic:

I. voiceless (aspirated?): p(h)

II. ejective: p'

III. voiced: b

These series had the following reflexes:

PIE series PIE (traditional) PIE (glottalic) Proto-Germanic Armenian Latin Greek Old Church Slavonic Sanskrit
I *p *p(h) *f f p p p p
II *b *p' *p p b b b b
III *bɦ *b *b b f-, -b(-) ph

Series I lenited in Proto-Germanic and to a lesser extent in Armenian:

*p(h) > f

*k(h) > s


*t(h) > th (not θ)

*kw(h) > kh (not x)

The aspiration of Armenian th and kh could be conservative.

II is no longer an ejective series in most modern IE languages*. Perhaps there was a chain shift in Germanic and Armenian:

*p' > *p > PG *f, Armenian f

Series I lenited, leaving a gap to be filled by II:

II > I > I'

In many branches of IE, II became voiced implosives (a different kind of glottalic consonant) before losing their implosive quality:

*p' > > b

(But is there a known example of ejectives becoming implosives in a non-IE language? Did the known shift *p- > ɓ- in Vietnamese have an ejective intermediate stage *p'-? I doubt it since I've never seen ejectives in Southeast Asian languages. Are there any known cases of ejectives directly becoming voiced obstruents?)

III remained voiced in Germanic and Armenian.

The devoicing and aspiration of III in Classical Greek has parallels in Thai, Lao, and many Chinese languages (e.g., Hakka):

*b > (*bɦ > *pɦ >) ph

However, those Asian languages did not lenite their aspirates:

Classical Greek ph > modern Greek f

(but ph and f coexist in Thai, Lao, Hakka, etc.)

The aspiration of III in Sanskrit has parallels in Shanghainese:

*b > (romanized bh)

The word-initial fricatives in III in Latin are from lenited, devoiced stops:

*b- > *bɦ- > *β- > f-

*d- > *dɦ- > *ð- > f- (cf. the Hawaii pronunciation of bathe with [f] instead of [ð])

*g- > *gɦ- > *ɦ- > h- (cf. *g- > [ɦ] in Czech, Slovak, and Ukrainian)

*gw- > *gwɦ- > *ɣw- > f- (cf. *gw- > f- in Hakka)

*Nonstandard varieties of Armenian preserve the ejectives. Dum-Tragut (2007: 17):

In some publications, the voiceless plosives [of Armenian] are also defined as ejectives or glottalised. Glottalised plosives occur in various Armenian dialects and can also be found in the Eastern Armenian vernacular based on the Yerevan dialect, but according to normative grammars, SMEA [Standard Modern Eastern Armenian] shows no glottalised voiceless plosives. THE MYSTERY MAN OF TRUE PEARL BAY

In Japanese and Korean, 'Pearl Harbor' is 眞珠灣 (modern Jpn spelling: 真珠湾) 'True Pearl Bay' rather than 珠港 'Pearl Harbor'. 灣 normally has wan-type readings in the Sinosphere: e.g., Mandarin wan, Cantonese waan, Sino-Japanese wan, all going back to Middle Chinese *ʔwæn which in turn is from Old Chinese *ʔron. But in Korean it is read 만 man and in Vietnamese it is read loan. These readings are presumably by analogy with those of other sinographs sharing the same phonetic:

蠻 Sino-Korean man < MC *mæn < OC *mron 'southern barbarian'

The 東國正韻 Tongguk chŏngun dictionary (1448) lists the idealized, regular Sino-Korean reading ʔwan and 새字典 Sae chajŏn lists an alternate modern reading 완 wan that I've never seen in use.

鸞 Sino-Vietnamese loan < MC *lwan < OC *ron 'mythical bird'

But 灣 is probably more common than those graphs. Why read a common graph (灣) like uncommon graphs? Are the m- and l-readings hypercorrections or based on lost readings in earlier Chinese dialects?

The unusual Korean and Vietnamese readings of 灣 are also in the K and V names for 臺灣 Taiwan: K Taeman and V Đài Loan (cf. Md Taiwan, Ct Toiwaan, SJ Taiwan).

12.9.1:40: I think I learned 湾 (the modern Japanese simplification of 灣) first, then 蛮 (the modern Japanese simplification of 蠻) and finally 鸞 (so infrequent in Japanese that it has no official simplification). In Chih-hao Tsai's list of sinographic frequency in Chinese:

灣 is #213 (followed by 月 'moon' #214)

蠻 is #830 (a lot higher than I expected; not considered important enough to be in the Oxford Starter Chinese Dictionary; 1/5 as frequent as 灣 which is in the Starter; 蠻 is surprisingly also among the 1,800 graphs taught in South Korean schools and the 1,945 graphs taught in Japanese schools, though it's in the least common of Morohashi 1992's categories of frequency)

鸞 is #4266 (1/135th as frequent as 灣; out of other l-graphs with the same phonetic [巒鑾臠欒曫羉圝灓孌孿攣], the most frequent is 巒 #3517 with 1/296th the frequency of 灣 and most of the others are rarer than 鸞 or even absent from Tsai's list) WHERE CLOTH CRIES

Today I was looking at the coverage of the Pearl Harbor attack in the December 9, 1941 issue of the Korean newspaper 每日申報 Maeil shinbo: e.g.,


Hawai (?)-ŭi Mi hamdae-rŭl kisŭp

'Surprise Attack on US Fleet in Hawaii' (see here for a literal translation*)

I'm not sure how 布哇 was read in Korean. It is an obsolete spelling for Japanese Hawai 'Hawaii' now written only in katakana as ハワイ. I presume it was also read as 하와이 Hawai in Korean. The logic behind the spelling has eluded me for twenty years. The normal readings of 布 'cloth' are fu and ho, not ha, and the most common reading of 哇 'to vomit (!); sound of crying' is ai, not wai. (Dictionaries list other readings - a, e, wa, e < we - but I haven't seen them in use.)

Common wai-graphs would have been better phonetic matches, but they have negative meanings that would make them poor choices for spelling a name associated with paradise: e.g.,

歪 'crooked'

猥 'obscene'

賄 'bribe'

矮 'dwarf'

However, there are many possible ha-graphs without such semantics -

葉 'leaf'

羽 'feather'

波 'wave'.

- so there was no need to choose 布 fu/ho for Ha-. The vocalic mismatch reminds me of 花瑠瑠 Honoruru for Honolulu with 花 hana corresponding to Hono-. (Could 花 reflect the pronunciation of Honolulu as [hanalulu] by some English speakers?)

The following article title could be from a modern North Korean newspaper if it didn't contain the Chinese characters no longer used there:


Kun-ŭl chŏltae shilloe hara!

'Have absolute faith in the [Japanese] Army!'

賴 has a strange Sino-Korean reading 뢰 roe (loe after -l). I would expect it to be 래 rae without rounding since nearly all other readings lack rounding: e.g., Mandarin lai, Cantonese laai, Sino-Japanese rai, Sino-Vietnamese lại). One exception is colloquial Taiwanese lua which may be from *lɔ < *lah < Old Chinese *rats. It is highly unlikely that Korean borrowed 뢰 roe from a Southern Min language related to Taiwanese. I can't think of any other Sino-Korean reading with such rounding: e.g., colloquial Taiwanese 大 tua < *dɔ < *dah < Old Chinese *lats 'big' corresponds to Sino-Korean 대 tae, not 되 toe.

12.8.22:57: Moreover, the Southern Min reading of 賴 never had a final glide unlike the unknown source of Sino-Korean 뢰 roe < *roy.

The 東國正韻 Tongguk chŏngun dictionary (1448) lists the idealized, regular Sino-Korean reading 래 rae for 賴. Does the irregular current reading 뢰 roe reflect lost some Chinese dialectal form like *loj, *lɔj, or *lɒj? The rounding is not the result of a Korean sound change: Korean *-ay did not become -oe.

*This is a very loose translation. A very literal translation would be 'Surprise Attacks US Fleet of Hawaii'. The subject (Japanese forces) is unspecified in the original. ESSENTIALLY UNRECOGNIZABLE

Colonization can bring completely different languages in contact: e.g., French and Vietnamese. Although Vietnamese borrowed massively from Chinese during Chinese rule, borrowings from French are rare and heavily disguised by their Vietnamized phonology.

In Nguyễn Đình Hoà's 1966 Vietnamese-English dictionary, (ét-)xăng and (ét-)săng (both 'gasoline') are derived from French essence [ɛsɑ̃s].

Vietnamese favors monosyllabic roots, so I presume essence was initially borrowed as disyllabic ét-(x/s)ăng and then reduced to monosyllabic (x/s)ăng.

The borrowing of [ɛ] as ét rather than e reminds me of how English short [ɛ] is borrowed as e plus echo consonant in Japanese:

Eng essence [ɛsɛns] > Jpn エッセンス essensu (not esensu).

Vietnamese syllables cannot end in -s, so -t is the only possible coda that is close to the following consonant x/s-.

Vietnamese x is [s] and s is [ʂ] in the south but [s] in the north. I suspect that the spelling săng was coined by a northerner who pronounced both x and s as [s]. I don't know which spelling is older. Both x- and s-spellings coexist in the online version of Bui (1992) as well as Nguyễn (1966), but only the x-spelling is in Nguyễn's 1983 English-Vietnamese dictionary, the online version of Hyde (2008), tratu.vn, Wikipedia, and Wiktionary. I conclude that the s-spelling is obsolete. The s-spelling had at least two problems: (1) it would be read as [ʂaŋ] by southerners with a [ʂ] far from the original French and (2) it shared a spelling with the unrelated native word săng 'coffin'.

Vietnamese has no nasal vowels, so ăng [aŋ] is the closest possible approximation of [ɑ̃].

French final [s] corresponds to zero in Vietnamese, since Vietnamese does not allow final consonant clusters like [ŋs]. RIVER UNCLE AND HEAVEN FISH

I am puzzled by the spelling 天邪鬼 'heaven evil demon' for amanojaku. Although 鬼 (normally ki in modern Sino-Japanese) was once read as クヰ kwi, it isn't read as ku in any other word I know of. Are there other 鬼 ku-words?

The Japanese Wikipedia entry for amanojaku says,


'Can also be written as 'river uncle' and 'sea (?)'.'*

Ama and jaku are possible pronunciations of 海 and 若 (with an implied no between them), but can 河伯 (normally kahaku) really be read as amanojaku?

Another unusual spelling is 天逆毎 'heaven oppose every' for amanozako. 逆毎 (which looks like gyakumai or sakagoto) is never read zako anywhere else, and zako normally means 'small fry' (spelled 雑魚 'miscellaneous fish'). What did zako originally mean in amanozako?

*12.6.1:36: 河伯 and 海若 are originally Chinese terms for the god of the Yellow River and a sea god.

若 has many meanings but none make sense in 海若, so I wonder if 若 is a loan graph for (a non-Chinese word?) *nak.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2010 Amritavision