Variant sinographs can be useful for reconstruction. If a word X is written with phonetic elements A and B, one could - possibly erroneously - conclude that A and B sounded like each other. The problem is determining the degree of likeness.

好 Old Chinese *huʔ-s 'love' has variants with 丑 'ox; second of the twelve Earthly Branches' instead of 女 'woman' or 子 'child'. (The 丑-variants of 好 also represent a very rare [and obsolete?] surname according to Shuowen.) The reconstruction of 丑 'ox' is unclear. It would be simplest to derive its Middle Chinese reading *ʈhuʔ from OC *thruʔ, but other evidence points in different directions:

1. 丑 appears as a phonetic in nasal-initial graphs: e.g., 杻 Middle Chinese *ɲuʔ 'a kind of tree', implying that the *ʈh- of 丑 MC *ʈhuʔ is from OC *hnr- (< ?*r-hn-) rather than *thr-.

2. The twelve Earthly Branches were borrowed into Tai. Many are still recognizable, but 'ox' has an initial labial: e.g., Ahom plao. Although it's possible that Tai borrowed only eleven of the Earthly Branches, I regard 'ox' as a Chinese loanword. Tai pl- could be a simplification of *pn- (absent from Proto-Tai) in a southern late OC form *pnaw < *pʌ-nuʔ, whereas MC *ʈhuʔ < OC *r-hnuʔ.

3. Schuessler (2007: 191) linked 丑 to Mon-Khmer words with nonlabial-l-sequences: e.g., Proto-Viet-Muong *c-lu (> Viet trâu 'buffalo'). One could propose an areal root *lu with a labial prefix in Tai but nonlabial prefixes in MK. However, if the Tai p(l)-forms are from Chinese *pn-, the shared -l- in Tai and MK could be coincidental.

If 好 was written with phonetic elements 子 (! - see here) and 丑,  did 子 and 丑 ever sound like each other?

I am struggling to reconcile the *h- < ?*ʔʌ-hl- of 好 *huʔ-s 'love' or the *l- in other 子-graphs with the *r-hn- of 丑. I know of no other cases of *(h)l- and *hn- written with the same phonetic, and the case for *-l- in 好 is weak. Pulleyblank (1991: 55) reconstructed *xʷr- for 丑, but how does a velar fricative end up becoming a retroflex stop?

Next: A heavenly reconciliation.

(*More precisely, phonetic elements in Chinese characters.)

Last night, I mentioned that I saw the Chinese characters for 好 'good' and 奴 'slave' (!) on the table of a 'Japanese' restaurant, but I didn't have time to discuss their structure, so I'll do that here.

Both 好 and 奴 contain the element 女 'woman'.

One might think 奴 'slave' is 'female slave', but in fact there is a separate character for female slave' with the same left side: 婢.

In Old Chinese *na 奴 'slave', 女 *nraʔ < ?*rɯ-naʔ 'woman' is phonetic and 又 'hand' on the right is semantic. Proposed cognates within Old Chinese are 女 *nraʔ 'woman' and 努 *naʔ 'exert', but these words are more phonologically complex than 奴 *na 'slave' and it's unlikely that affixes were added to 'slave' to generate the more basic word 'woman' (though 'slave' > 'exert' [what slaves do?] might be possible).

In 婢 *beʔ < ?*N-pe-ʔ .'female slave', 女 'woman' is semantic and 卑 *pe 'lowly' is semantic/phonetic. 'Female slave' is probably derived from 'lowly' via affixation: *pe 'lowly' > *N-pe-ʔ 'the lowly one' > 'female slave'.

好 *huʔ 'good', *huʔ-s 'love' (< 'consider good'; Schuessler 2007: 273) is often considered to be a semantic compound of 女 'woman' with 子 'child', but Boodberg proposed in "Proleptical Remarks on the Evolution of Archaic Chinese" that 子 'child' is in fact phonetic. This seems unlikely on the basis of modern readings of 好 and 子 which don't sound remotely alike: e.g., Mandarin hao and zi. Boodberg proposed that 子 had two OC readings, *tsləg (> Md zi) and *KOG (no Md descendant), and that the second reading underlay 好 (his OC *xôg).

Decades later, Boltz (1994: 110) proposed that the following characters contain 子 as phonetic and represented cognate words:

(The sinographs 保毓育 once contained 子, though this is not apparent from their modern forms.)

In Boltz' reconstruction, all eight share the rhyme *-əkw/-əgw, but in mine (based heavily on Sagart 1999), all share *-u, and perhaps all of them can be traced to a common archetype *lu. However, there is no nongraphic evidence for *-l- in the words written as 孚保好孝. In other words, if I did not know that 孚保好孝were (or had been) written with 子, I would have no reason to reconstruct *-l- to make their readings similar to the *l-readings of 毓育斿汓.

The semantics of 斿汓 don't fit the other six, so I doubt they are cognate even if 子 is phonetic in all eight.

Tonight I had dinner at a Chinese-run 'Japanese' restaurant. As if the Japanese name that didn't even match the Chinese character in the logo for the restaurant weren't bad enough, the table was decorated with a pattern consisting of Japanese dictionary entries for 好 'good' and ... 奴 'slave'!?

That was tasteless, though the not-so-Japanese food tasted OK. I would have expected Chinese staff to have known better. It's creepy to eat from a plate surrounded by 奴 'slave'. I prefer the Korean Bible quotes in a Korean restaurant I visited in Oakland back in 1991.

09.7.8.22:44: KOREAN ROMANIZATION VERSION 789

(789 = July 8, 2009. Too bad I'm not writing this at 12:34:56!)

The question of how to romanize Korean will never be answered to everyone's satisfaction. The sound system of Korean simply doesn't map very well onto the Roman alphabet without diacritics. Nonetheless, here's how I would tackle the problem. This proposal only barely begins to deal with the many issues involved.

Korean has four types of stops and affricates:

In Korean spelling, the voiceless unaspirated and voiced series are written with the same letters: e.g., the voiceless unaspirated P- of 부산 Pusan is written with the same letter ㅂ as the voiced -b- of 사부 sabu 'master'. In the current South Korean official 'Revised Romanization', 부산 Busan and 사부 sabu are both spelled with b even though they are not pronounced identically. This is not a problem if one knows that b- in Revised Romanization is [p] word-initially, but it is unreasonable to expect most people to know that. Hence I think the distinction should be indicated in the romanization.

To non-Korean ears, voiceless unaspirated p may sound like a compromise between p and b, so b in Revised Romanization is not an entirely unreasonable choice. One could try to reflect this intermediary nature of p by romanizing it as pb or bp, but such combinations are awkward and would also be frequent.

I indicate aspiration with an h added to the symbols for the voiceless unaspirated series. An h is harder to ignore than the apostrophe used in the McCune-Reischauer romanization. The only problem is that one might think ph th kh were fricatives [f θ/ð x] though they are actually stops.

The three-way distinction of p/ph/b, t/th/d, k/kh/g is not unlike the Latin transcription of the same three-way distinction in Classical Greek:

The reinforced series (which has no counterpart in Greek or any European language) is indicated with doubled consonants except for tch (carried over from McCune-Reischauer) instead of chch.

The problem of the 'extra' vowels

Korean has five vowels a e i o u that map nicely onto the five vowel letters a e i o u.

However, it also has at least three vowels that don't: ɛ ʌ ɯ. These sound almost like e o u (and some speakers pronounce e and ɛ identically), so for general purposes, I would romanize them as e o u. In more precise romanization, I would add circumflexes to distinguish them from regular e o u: ê ô û. This situation would be analogous to Vietnamese which also uses diacritics to distinguish vowel qualities, though these diacritics are usually left out by non-Vietnamese. I prefer not to use the breves of McCune-Reischauer since breves normally indicate short vowels and ɛ ʌ ɯ can be short or long. (Long vowels can be optionally doubled in the most precise version of my romanization).

I don't favor digraphs for single vowels because they look like vowel sequences:

McCune-Reischauer and Revised Romanization ae for [ɛ] looks like [ae]

Revised Romanization eo for [ʌ] looks like [eo]

Revised Romanization eu for [ɯ] looks like [eu]

Comparison of romanizations

09.7.7.23:59: THE LANGUAGES OF HADANUS: URDREH NOUNS FROM VERBS

I recently considered having Urdreh imperatives be

root + number suffix + gender suffix

e.g.., mem-y-e! 'eat!' (said to a single male)

However, I forgot that

root + number suffix + gender suffix + zero

is the structure of Urdreh vocatives, so mem-y-e! would mean 'eater!' (sg. masc.). To avoid overlap with vocatives, I now consider bare verb roots to be imperatives: mem! 'eat!'

Urdreh roots plus number, nonabstract gender, and case endings (including the zero of the vocative) form agent nouns:

mem-y-e-h 'eater' (sg. masc. absolutive)

mem-w-a-h 'two eaters' (dual fem. absolutive)

mem-l-u-h '(three or more) eaters' (countable plural concrete neut. absolutive)

The abstract neuter ending forms action nouns:

mem-y-i-h 'the act of eating' (sg. abst. neut. absolutive)

In The Seeding, kah is defined as 'hated', and more recently, ka (no h) was defined as 'hate'. This implies that -h- forms past passive participles. -h- is added directly to vowel-final roots:

ka 'hate' > kah-yeh 'male Kasgen' (lit. 'he who is hated')

cf. ka-yeh 'hater' (masc.)

ji 'lead' > jih-reh 'followers' (lit. 'those who are led'; masc.)

cf. jeh < ji-yeh 'leader' (masc.)

ma 'do' > mah-yuh 'that which was done'

cf. ma-yah 'doer' (fem.)

na 'parent, give birth' > nah-reh 'children' (lit. 'those who were born'; masc.)

cf. na-yah 'mother' (lit. 'female parenter')

To avoid complex clusters with h, I propose that 'echo' vowels are inserted between root-final consonants and -h-:

il- 'run' > il-ih- 'ran' > il-ih-yuh 'that which was ran upon'

cf. il-yeh 'runner' (masc.)

mem- 'eat' > mem-eh- 'eaten' > mem-eh-yuh 'that which was eaten'

cf. mem-yeh 'eater' (masc.)

kad- 'talk' > kad-ah- 'talked' > kad-ah-yuh 'that which was talked about'

cf. kajah < kad-yah 'talker' (fem.)

drey- 'push' > drey-eh- 'pushed' > drey-eh-yuh 'that which was pushed'

cf. drey-ah 'pusher' (fem.)

The verb root plus -(V)h can be added directly to a noun to form a compound: e.g.,

mem-eh-sah-yuh 'the food that was eaten'

equivalent to a past passive participle + noun phrase

mem-eh-yuh sah-yuh 'the food that was eaten'

Long compounds of short past passive participles, adjectives, and nouns are possible:

mem-eh-yus-su-sah-yuh 'the green [lit. yellow-blue] food that was eaten'

09.7.5.21:21: THE LANGUAGES OF HADANUS: COMMON URDREH VERBS

Urdreh verbs have the following structure:

There is no distinction between noun and verb roots: e.g., ji- < di- 'lead' in Jeh < Ji-yeh < Di-yeh is also a verbal root in forms like

jeba < ji-ye-ba 'thou led' (with the same irregular contraction of jiy- > j- as Jeh)

Jiler Urdreh jilebe 'the (three or more) Jeh lead the (innumerable) Urdreh'

(word order like Japanese here for the time being; actual Urdreh word order is unknown)

Number and gender refer to the number and gender of the subject.

Urdreh person suffixes were meant to be as un-Indo-European as possible. Sanskrit exemplifies the old Indo-European pattern:

as-mi 'I am'

asi < as-si 'thou art'

as-ti 'he/she/it is'

I deliberately chose Urdreh person markers that were

- not nasals (to be unlike -mi)

- not fricatives (to be unlike -si)

- not voiceless stops (to be unlike -ti)

- not at the same points of articulation as -mi, -si, -ti (1st labial, 2nd and 3rd alveolar)

Urdreh distinguishes between 1st exclusive -d- ('we, not including the listener') and 1st inclusive -rb- ('we, including the listener') < -d-b-.

Mood suffixes presumably use consonants not already used for number or person followed by a vowel (to distinguish them from noun/adjective case endings which consist of single consonants).

Examples using the verb stems Robinson posted yesterday (note dy > j, di > ji as in Japanese):

il-ya-da 'I (female) ran'

mem-ye-de 'I (male) eat'

normally not used; see below

kajudu < kad-yu-du 'I (neuter concrete) will talk'

a talking object in a children's story might say this

dreyyiji < drey-yi-di 'I (neuter abstract) 'I (neuter abstract) would push'

the personification of an abstraction might say this in a myth

yala-wa-ba 'you two slept'

yala is disyllabic and may be a compound of two roots ya-la- ('close-eye'?)

ma-la-rbe 'we (inclusive; you and I; three or more people) do'

ka-ra-gu 'they (large number of people) will hate'

Urdreh typically avoid first person singular forms and prefer inclusive forms: e.g.,

mem-ya-du 'I (female) will eat'

sounds like something a lone Urjah (< Urd-yah; single female Urdreh) would do if all alone (a state dreaded by Urdreh) or like a translation of Kasgen or Dwar first-person speech. If there are two female Urdreh (Urdwah) in a room and only one is eating, the eating Urjah would say to the non-eating Urjah,

mem-wa-rbe 'thou and I (both female) eat'

The non-eating Urjah would feel included even though she is not actually eating. The -rb- forms are hence much more common in Urdreh than 'you and I' or 'we' (inclusive) are in English.

If Urdreh are not getting along, they can express hostility by excluding the other and not using -rb-forms: e.g.,

mem-le-de 'we (masc. exclusive countable pl.) eat' (and we don't care to include you with us)

mem-le-rbe 'we and you eat'

Next: Deriving Urdreh nouns from verbs.