07.9.1.20:36: A LAT OF DERIVATIVES

So far, I've been talking about Chinese words with *-p ~ *-k alternations implying an earlier *-kw. Final *-p can also alternate with *-t within members of a phonetic series. Let's look at Karlgren's series 339 and 633 (definitions are not exhaustive; a,e = vowel may be *a or *e):

339a-c 世 MC *shieyh < OC *hlaps 'generation, epoch; world'

339d 貰 MC *shieyh < OC *hlaps, *jah ~ *zhah < OC *m-laps 'lend, borrow; remit'

339e 抴 MC *yieyh < OC *l(a,e)t-s, MC *yiet < OC *l(a,e)t 'to pull; oar'

339f 枻 MC *yieyh < OC *l(a,e)t-s 'oar'

339g 詍 MC *yieyh < OC *l(a,e)ts 'garrulous'

339h 泄 MC *siet < OC *s-hl(a,e)t, MC *yieyh < OC *l(a,e)t-s 'leak out'

339i 紲 MC *siet < OC *s-hlet 'to bind' (homophonous with 'leak out'!); also used to write 330n 褻 OC *s-hnget 'garment next to the body' (in a dialect in which *hl and *hng merged?)

339j 鞢 MC *siet < OC *s-hlet 'bridle strap' (cognate with 339i; something that binds?)

339k 勩 MC *yieyh < OC *laps, MC *yih < OC *ləps 'toil, fatigue'; also written as 509g 肄 MC *yih < OC *ləps 'toil' (whose apparent phonetic 聿 OC *lut is a corruption of 隶 OC *ləps ~ *ləps 'come to', *leps 'peaceful')

339l 渫 MC *siet < OC *s-hl(a,e)t 'cleanse; leak'

339m 緤 MC*siet < OC *s-hlet 'bind' (cognate with 339i, 339j)

633a-c 枼 MC *yiep < OC *lap 'generation'

633d 葉 MC *yiep < OC *lap 'leaf'

633e 鍱 MC *yiep < OC *lap 'to plate with metal'

633f 堞 MC *dep < OC *lep 'parapet'

633g 牒 MC *dep < OC *lep 'tablet'

633h 蝶 MC *dep < OC *lep 'butterfly'

633i 褋 MC *dep < OC *lep 'unlined garment'

633j 諜 MC *dep < OC *lep 'spy'

633k 蹀 MC *dep < OC *lep 'trample'

633l 揲 MC *yiep < OC *lep, MC *sep < OC *s-hlep, MC *zhiet < OC *m-let, MC *shiet < OC *hlet 'take and measure'

633m 偞 MC *yiep < OC *lep, MC *xiep < OC *'-hlep 'small, insignificant'

633n 亻+ㄊ+木 (variant of 633m with abbreviated phonetic)

633o 韘 MC *shiep < OC *hl(a,e)p 'archer's thimble'

633p 屧 MC *sep < OC *s-hlep 'bottom inlay in shoe; shoe'

339a-c 世 is phonetic in graphs for syllables with four types of codas:

*-p: 枼葉鍱堞牒蝶褋諜蹀揲偞韘屧

*-ps: 世貰勩

*-t: 抴泄紲鞢渫緤

*-ts: 抴枻詍泄

The majority have *-p(s), and 葉 OC *lap matches perfectly with Proto-Tibeto-Burman *lap 'leaf' (as reconstructed by Matisoff 2003: 656). This indicates that 世 originally represented *lVp(s)-syllables, and that the *lVt(s) cases are exceptional.  Could*-t(s) have derived from some earlier cluster like *-pt(s)? Perhaps, but the presence of such a cluster would also imply the existence of other final stop-stop clusters such as *-kt, and I know of no evidence for them.

I suppose I could explain the *-p ~ *-k alternations in the phonetic series for 立 by reconstructing final clusters like *-kp, but I'd rather reconstruct *-kw whose existence in OC is already certain.

The key to understanding *-p ~ *-t alternations is knowing that some dialects of OC had merged *-ps and *-ts into *-ts (or at Starostin reconstructed it, *-ch). Here's what may have happened:

Scenario A:

2. OC *-ps and *-ts merged into *-ts.

3. This resulted in two kinds of alternations:

*-t ~ *-ts (< earlier *-t ~ *-ts)

*-p ~ *-ts (< earlier *-p ~ *-ps)

Hence an earlier 

抴 OC *l(a,e)p ~*l(a,e)p-s 'to pull; oar'

became a later

抴 OC *l(a,e)p ~*l(a,e)ts 'to pull; oar'

4. *-p words were replaced by *-t words by analogy with *-t ~ *-ts alternations: e.g.,

Given that the unsuffixed version of

說 OC *hlots 'exhort' (< orig. *hlot-s)

ended in *-t:

說 OC *hlot 'speak' (< orig. *hlot)

it would not be unreasonable for an OC speaker to guess that the unsuffixed version of

抴 OC *l(a,e)ts 'to pull; oar' (< orig. *l(a,e)p-s)

should also have ended in *-t:

抴 OC *l(a,e)t 'to pull; oar' (replacing orig. *l(a,e)p)

Scenario B assumes the reverse. If the*-t(s) words in question (抴泄紲鞢渫緤枻詍) were first written after the shift of *-ps to *-ts, then speakers would use *-ts (< orig. *-ps, *-ts) phonetic elements to write both *-p(s) and *-t(s) words.

Given that

324a-c 兌 OC *lots 'glad; open a passage'

was used as phonetic in graphs for *-t(s) words like

324e 蛻 OC *lot ~ *hlot-s ~ *hlot-s ~ *hlor-s (with *r!) 'exuviae of insects or reptiles'

324j-k 敓 OC *lot 'snatch'

324l 挩 OC *lot ~*hlot 'take away; beat, kill' 

324m 脫 OC *lot ~ *hlot 'peel off'; *hlot-s 'leisurely'

324n 梲 OC *tlot 'short pillars'

324o 悅 OC *lot 'pleased'

324p 閱 OC *lot 'hole; inspect; count'

324q 說 OC *hlot 'speak', *hlot-s 'exhort'

it would have made sense to use

世 OC *hlats (< orig. *hlaps) 'generation, epoch; world'

as phonetic in graphs for *l-t(s) words (抴泄紲鞢渫緤枻詍).

- how old the *-t(s) words (抴泄紲鞢渫緤枻詍) are

- when the *-ps > *ts shift occurred

A third scenario would involve some lost conditioning factor that triggered a *-p(s) to *-t(s) such as a labial-initial prefix:

*pə-lap(s) > *lat(s)

(9.2.00:02: Cf. the *-p > *-t shift in Cantonese

法 *faap > faat

to avoid a labial onset and a labial coda in the same syllable.)

07.9.5.00:23: Schuessler (2007: 537) proposed that 泄 (my OC *s-hl(a,e)t ~ *l(a,e)t-s) might have had *y in OC. However, I don't understand why a *y-initial word would be written with an *hl-initial phonetic (世 - my OC *hlaps and Schuessler's OC *lhats < *-ps).

Schuessler links 泄 (with a possible root ?*yat in his OC) with Mru yat 'to leak, ooze'. He wrote,

The phonetic [世] implies an OC L- or J- [i.e., y-] like initial, Mru's initial y- seems sometimes to correspond to PTB [Proto-Tibeto-Burman] *l- (beside *j- [y]; it certainly corresponds to PTB *l- in ya 'easy' [cf. 易 OC *leks 'easy'], therefore the root initial in this group is not certain.

Even if the initial y- in Mru yat is from an *l- that does match the lateral initial implied by the phonetic 世 OC *hlaps, the problem of the non-matching codas remains. 世 implies an origiinal root-final *-p, not a*-t which would match Mru -t. Is this a case of converging unrelated forms?

early OC *s-hl(a,e)p ~ *l(a,e)p-s > later OC *s-hl(a,e)t ~ *l(a,e)ts resembling pre-Mru *?lat

07.8.31.23:56: A 世 WORLD OF 火 FIERY 木 TREES

Before I go on vacation, I'd like to clarify what I meant by going "back into the frying pan".

I'm still looking for members of the Old Chinese *l-kw 'bright' word family. Meet

煠 MC *yiep

defined as its possible cognate 爚 MC *yïak < OC *lakw 'shine'.

Although 煠 is not attested in OC, it can't have come out of nowhere. The probable OC ancestor of MC *yiep would have been something like *lep (< ?*lekw) which is not far from 煜 OC*w-ləp (?*w-ləkw) ~ *luk (?*ləkw) 'gleaming'.

煠 had four more MC readings.

Its second MC reading *ChhEp [ʈʂhɛp] with a retroflex affricate initial implying OC *r-s-t-hlep (< ?*r-s-t-hlekw) was defined as 爚 MC *yïak < OC *lakw 'shine'. The functions of the prefixes are unknown.

Its third and fourth MC readings

*ThEp [ʈhɛp] < OC *r-t-hlep (< ?*r-t-hlekw)

*zhEp [ʑɛp] < OC *m-lep (< ?*m-lekw)

(the combination *zhE is unexpected in MC; was the fanqie 實洽 *zhit-GEp supposed to represent an expected *zhiep?)

were defined as 瀹 MC *yïak < OC *lakw 'drain off; clear the course (of a river); purify (the heart)' (i.e., make shine?). Once again, the prefixes are obscure.

Interestingly, 煠 MC *ChhEp 'shine' had an MC homophone written as 竹 'bamboo' plus the same phonetic [世+木] defined as 籥 MC *yïak < OC *lakw 'flute' (cognate to 笛 OC *likw 'flute'). Note the parallelism which is not likely to be coincidental:

煠 < OC ?*lep def. as 爚 MC *yïak < OC *lakw 'shine'

煠 < OC ?*rthlep ~?*mlep def. as 瀹 MC *yïak < OC *lakw 'drain off ...'

[竹 +世+木] < OC ?*rsthlep def. as 籥 MC *yïak < OC *lakw 'flute'.

It might be possible for one *l-p word to be defined as an unrelated *l-kw word. But how likely is it for four *l-p words to be defined as unrelated *-kw words? Were 煠 and [竹+世+木] nonstandard cognates of 爚, 瀹, and 籥 which were not written until the MC period?

The fifth MC reading of 煠 was *JEp [ɖʐɛp] (perhaps from OC ?*n-r-s-t-lep) defined in Guangyun as a non-cognate, 湯 *thang < OC *hlang 'hot water'. Nearly a millennium later, Lin Yutang's dictionary defined 煠 as a variant of 炸 'fry in deep oil', a word postdating the MC period. Mandarin zha and Cantonese ja for 炸 imply an MC *Chæh [ʈʂæh] which could have come from an OC ?*r-s-t-lakw-s.

I'll try to be back on Tuesday.

9.1.00:27: You may be wondering how an OC word can have as many as four prefixes: e.g., 煠 ?*n-r-s-t-lep. This analysis is required by Sagart's (1999) OC reconstruction which only allows for monoconsonantal prefixes. I wonder if *s-t- (possibly *t-s-) was really an affricate prefix *ts- or a reduction of an earlier monosyllabic word *sVt or *tVs.

07.8.30.23:56: TO ME OR NOT TO ME

While reading about sign language to write last night's post, I found this passage:

Meier 1990 demonstrates that only two grammatical persons are distinguished in ASL: First person and non-first person, as in Damin.

Until I read that, I had never questioned the assumption that a language had to have at least three grammatical persons.

However, I should point out that Damin is a "ceremonial language register", not the normal language of an entire community. Does any community of oral language speakers normally use only two grammatical persons?

I think I have heard of fourth person before, but I know almost nothing about it, and I am totally ignorant about categories with even higher numbers: e.g., fifth person. I do know much more about sinography, so I'll return to that topic soon.

07.8.29.23:59: TYPE 5 GRAPHS AND THE SOUND OF SILENCE

In "火 Fiery 羽 Feather 立 Stance", I asked,

Where are the type 4 G?

They're all over this website. They're letters. They are [+P -S]: they have conventionalized phonetic values but have no conventionalized semantic values.

Not all letters are type 4. Some are type 3 when used to write words (i.e., when they are [+S] as well as [+P]): e.g.,

a (indefinite article)

B (abbreviation of be)

I (first person singular pronoun)

O (in vocative phrases: O God)

Q (name of a character in Star Trek: The Next Generation)

U (abbreviation of you)

V (name of a US sci-fi franchise)

Y (abbreviation of why)

Any letter is type 3 when its S is itself: e.g., Z is [+P +S] if it is pronounced [zed] or [zi] and means 'the letter Z'.

The letters G, P, S are type 3 in these posts since they represent words as well as the sounds [ji] [pi] [es].

Are all letters type 3 or 4? What if a letter is silent? If it has no P, then it can't be type 3 or 4 - right? And if it has no meaning, then is it type 1 [-P -S]? Does the final letter of mute belong to a category of "merely idiosyncratic or random marks or drawings" (Boltz 2003: 19)? Is mute 75% writing (mut) and 25% non-writing (e)?

My answers are 'no' and 'no', because the 'silent e' of mute is not actually silent. The English letter sequence u ... e represents [yu] in mute [myut]. Removing the e would result in mut which might rhyme with but [bʌt]. e therefore has a P, though this P is not straightforward: it helps to specify a preceding vowel.

But what about truly silent letters like the final e of Lynne? Removing it has no phonetic effect: Lynne and Lynn are homophones. What differentiates this final e from a random squiggle? e has conventionalized phonetic values elsewhere in the English writing system, whereas a squiggle has no P and is not part of any system. I could describe the e of Lynne as [-S (+P)] - a fifth type of graph? That e could theoretically be read. Just as Bette is pronounced [bɛti] when referring to Bette Davis, Lynne could be read as [lɪni]. A squiggle, on the other hand, could not be read.

Letters that are 'even more silent' could exist. Suppose speakers of a language with the phoneme h invented an alphabet with a letter for h. Now suppose that the phoneme h completely disappeared from speech but spellings with the letter for h persisted. Has the letter for h become nonwriting because it lost its P? I could argue that it is [-S (+P)] because it once had a conventionalized phonetic value [h] which current speakers cannot pronounce. [-S (+P)] graphs are either potentially or previously phonetic. Either way, they have phonetic associations that distinguish them from graphic 'noise'.

No writing system has ever incorporated graphs that were always and only [-S -P]. Nobody has ever designed an alphabet with a silent letter from the outset. Moreover, no writing system consists largely or wholly of silent letters for an obvious reason - writing systems have to represent sounds. A 'silent writing system' would be an oxymoron. It would be a system, but it wouldn't be writing from a phonocentric perspective.*

I would like to regard silent letters as a subcategory of type 4 [-S +P] rather than as a separate fifth type. They were 'born' as type 3** or 4, but they lost their P later in 'life'. They are secondarily, not originally silent. Nonetheless, they are still part of a writing system; they combine with other letters to represent spoken words and are not meaningless, random add-ons.

Next: Back into the frying pan.

*8.30.00:27: P would have to be redefined in terms of cheremes*** for writing systems representing sign languages. The sign language equivalent of a 'silent writing system' would contain [-P] graphs which didn't correspond to a component of a sign.

If writing systems such as Stokoe notation or SignWriting became fixed while signs changed over time, some graphs might become secondarily 'silent' ('nonrepresentative'?).

**Type 3 graphs can lose their semantic value and become pure phonetic symbols (i.e., type 4 graphs).

***Although the term phonemes can also be used for sign languages, I use the term cheremes to avoid ambiguity.

07.8.28.23:59: 火 FIERY 羽 FEATHER 立 STANCE

I've been reading William G. Boltz' The Origin and Early Development of the Chinese Writing System (2003 ed.; Amazon sells the 1994 ed.). Boltz is arming me with the conceptual and terminological tools I need for better blogging about writing. And I haven't even gotten to his chapters on Chinese yet! (Well, I have read parts of those chapters out of order, but that doesn't really count.) His chapter "Writing in General" (p. 19) introduces three useful abbreviations:

G = graph

P = phonetic value

S = semantic value

Graphs can be classified into four types according to whether they have a P and/or an S:

1. G : [-P -S]

Type 1 G have no phonetic or semantic values. They are "merely idiosyncratic or random marks or drawings, and have no bearing on writing."

2. G : [-P +S]

Type 2 G "are visual signs that communicate meaning, but because they have no automatic and unambiguous relation to [spoken] language ... they are, by [Boltz'] definition, not writing."

For Boltz - and for me - a graph must be [+P] if it is to be writing. Boltz gives a skull and crossbones as an example of a type 2 G which

... could be "read" variously as 'poison', 'poisonous', 'hazardous', 'pirate', or even 'skull and crossbones'. Because of this linguistic variability, the skull and crossbones graph is an example of the communication of an idea [or a set of ideas -A] directly rather than one governed or mediated by language. On this basis we would deny it the status of writing. To do otherwise leads to chaos: we would have to admit as writing every image , painting, graphic symbol or icon that evoked a meaningful association in the mind of the beholder. (p. 19)

The icons on my computer screen are meaningful, but they aren't like the Chinese characters or Roman letters around them. They have no fixed set of 'readings', and they are not uniquely linked to a specific language or set of languages.

The remaining two types of G are [+P] and therefore are writing:

3. G : [+P +S]

4. G : [+P -S]

The title of this post describes a type 3 G: 熤. It had two P in Middle Chinese, *yiek and *yik, but had only one S: 'the personal name of 張熤 *Tïang yVk (who lived during the Northern Wei Dynasty)'. Apparently there was no consensus about how to read his name.

I suspect the name is just another spelling of 翌 MC *yik ~*yuk 'bright' (see here) with 火 'fire' added as a semantic element. Note that 'name' here refers to the specific G 熤. Although 翌 may be etymologically identical to 熤, it is not 張 *Tïang's personal name. And the P of 熤 and 翌 are not quite the same: the two share an MC reading *yik (< OC *lək[w]) but they also have mutually exclusive readings:

熤 MC*yiek (< implying OC *lek[w]; the name is not attested in OC)

翌 MC*yuk < OC *luk or *ləkw

The three readings may derive from a common OC root *l-k(w). Other derivatives may include this set of cognates proposed by Schuessler (2007: 477, 632).

爚 OC *lakw 'shine'

瀹 OC *lakw 'drain off; clear the course (of a river); purify (the heart)' (i.e., make shine?)

曜, 耀, 燿 OC *lakw-s 'be brilliant, shiny'

濯 OC *r-lakw 'be clean, brilliant, bright, glossy; to moisten, wash'

濯 OC *r-lakw-s 'to wash clothes'

(Ironically, 濯 and 濁 OC *rtok 'muddy' are now both pronounced zhuo in Mandarin.)

爍 OC *hlakw (< ?*s-l-) 'shine'

滌 OC *likw 'to clean'

I also suspect that the OC *l-k and*l-p (< *-kw?) words in these previous posts are cognates.

Next: Where are the type 4 G?

8.29.00:21: Is the OC *l-m 'blaze' word family (see here) also related to the *l-kw word family? Could *l-m ultimately be from *l-kw plus a nasal suffix?

07.8.27.23:09: THE 日立 HITACHI HINT

(The brand name Hitachi is written as 日立 'sun-stand' in Japanese.)

In "When Do 羽 Feathers 立 Stand?", I mentioned 煜 'illuminate, shine'. According to Shuowen (100 CE),

煜 means 熠 'gleaming' (Xu ed.; the Duan ed. defines it as 燿 OC *lekw 'shine', which could still be cognate)

the phonetic of 煜 is 昱 'sunlight'

the phonetic of 昱 is 立 'stand'

One would expect 煜, 昱, and 立 to sound alike. But the first two clash with the third:

As far as I know, neither 煜 nor 昱 have early textual attestations. This implies that the graphs were created at a later period when those words still sounded like 立. The three words may have been nearly homophonous even in Xu Shen's Late Old Chinese dialect, though it's possible that Xu Shen was simply transmitted received knowledge that was no longer relevant to his speech.

How can 煜 nor 昱 be reconciled with 立? Even if I derived *rəp from*rəkw, the root initials (*l and *r) would still not match. And I cannot think of any other cases of *l-words beting written with *r-phonetics.

There is one way out, but it comes with a price. The Middle Chinese reading *lip for 立 is pretty secure. Although its vowel is open for discussion, no one argues that it had an MC onset other than *l-, or an coda other than *-p. According to Sagart (1999: 19, 28), MC *l- has two sources, OC *r- and OC *Cə-l- (i.e., *l preceded by an unstressed prefix). So perhaps 立 MC *lip came from an earlier OC *Cə-ləkw, whose root would match the OC *ləkw I have reconstructed for 煜 and 昱. But if the OC word stand began with *l-, then it can no longer be linked to Tibeto-Burman words for 'stand' with r-. OC *l- corresponds to TB l-, not TB r-. Thus the vague similarity between the OC and TB words for 'stand' would be coincidental. The internal evidence of the script points one way (立 had OC *l like 煜 and 昱) and the external data tempts one to go the other way (立 had OC *r).

I can't make up my mind. On the one hand, I want to resist temptation and avoid the risk of positing a false pair of lookalike cognates. On the other hand, if I give in to temptation and link 立 to the TB *r-words, then I'll need to posit a semantic function for 立 'stand' in 昱 'sunlight'. And I can't think of any.

Next: Fiery feather stance.

07.8.26.23:49: WHEN DO 羽 FEATHERS 立 STAND?

Answer: the 翌日'next day'.

In my last post, I suggested that 立 OC *rəp 'stand' might have had a final labiovelar *-kw.

Such a velar final might explain why 立 OC *rəp 'stand' is in 翌 'buzz about (while flying); bright; next (day)'. If a graph XY means Z, and X has a semantic function, but Y does not, Y is probably phonetic. In the case of 翌, 羽 OC *wa 'feather' is semantic, so 立 'stand' should be phonetic (as suggested by Pulleyblank [1991: 51]).

The trouble is that most evidence does not indicate that 翌 ever sounded much like 立 OC *rəp 'stand'. 翌 has two MC readings, *yik and *yuk. These imply the OC readings *lək and *luk (or *ləkw?). Moreover, 翌 'next (day)' was also written as 翼 OC *lək 'wing', a compound of 羽 OC *wa 'feather' (semantic) + 異 *ləks 'strange' (phonetic). None of these words have *r or *p.

The phonetic differences between 立 OC *rəp 'stand' and 翼 OC *lək 'wing' also have parallels in Tibeto-Burman (a term I'll use loosely to mean 'Sino-Tibetan other than Sinitic'):

'stand'

Written Burmese rap (also 'stop, halt, remain')

Tangut TT2117 'ya^r 1.82 < ?*rya

'hand, arm'

Written Tibetan lag < *lak

Written Burmese lak

Tangut TT1545 lạ 1.63 < *C-la

(Data from Gong [1995: 79, 88]. The correspondence of Old Chinese *ə to Tibeto-Burman a is expected. No, I haven't forgotten Tangut! I plan to analyze those tangraphs later this week.)

However, one must not regard 立 OC *rəp 'stand' and 翼 OC *lək 'wing' as correct reconstructions simply because they resemble words in neighboring languages. It's possible that those reconstructions are valid for a stage of Chinese postdating the stage underlying the script.

It's tempting to regard 翌 OC *lək 'buzz about (while flying)' as an semantic extension of 翼 OC *lək 'wing'. Perhaps 翌 OC *lək had a lost presyllable (a derivational prefix?) differentiating it from the noun 翼 OC *lək.

Why did I write "most evidence"? It's because Pulleyblank (1984: 166) pointed out that 煜 'illuminate, shine' has two MC readings, *wïp and *yuk. 煜 is presumably related to 翌 MC *yik ~*yuk 'bright'. Moroever, Pulleyblank regarded 煜 as a variant of 熠 'gleaming' which had final *-p (details here). This *-p ~ *-k variation in words for brightness suggests a proto-labiovelar final *-kw which became *-p or *-k depending on preceding segments (and on the dialect?).

Wait ... I've been ignoring what might be the best evidence besides the script itself. But I'll save that for tomorrow.

Next: The Hitachi Hint.