on the subject of early Chinese pronouns ... for now. Not forever.

I was embarrassed last night when I wrote improbable forms like

√*ɢʷʕŋ 'go'

When I first came up with the zero grade idea, I reconstructed that root as a far less exotic

√*wŋ

I got the *ɢʷ from Sagart's uvulars paper. I no longer think the graph 于 'go' belongs to an Old Chinese uvular series, and I've gone back to the forms I would have reconstructed a week ago:

(Sagart's gloss 'impure' is for MC *ʔoh, another reading of 汙. Medial *-r- should not be optional in Sagart's reconstruction of the OC ancestor of MC *ʔwæ. His OC *qqʷa would become MC *ʔo.)

Although I believe OC had uvulars, I have even more doubts about Sagart's proposal than ever. OC had several *l-initial first person singular pronouns:

予 *Cɯ-la(ʔ)

(see Pulleyblank 1995:77 for the glottal stop)

余 *Cɯ-la

台 *lə

朕 *r(ɯ)-ləm

Sagart contrasted lateral series like 余 with uvular series such as

與 Sagart's OC *ɢ(r)aʔ 'give; for; and'

舉 Sagart's OC *Cə-q(r)aʔ 'lift; raise'

(Sagart presumably left out glottal stops by accident.)

which I would reconstruct as

與 OC *Cɯ-laʔ 'give; for; and'

舉 OC *kɯ-laʔ 'lift; raise'

Notice that 予 'I' and 與 'give' share homophonous roots in my reconstruction but not in Sagart's:

予 can also write an OC word for 'give' ending in a glottal stop.

I think that 予 and 與 represented the same root for 'give' (*laʔ) with different prefixes:

予 OC ?*sɯ-laʔ (cf. 序 *sɯ-la-ʔ 'school')

與 OC ?*kɯ-laʔ (cf. 舉 OC *kɯ-laʔ 'lift; raise')

My guesses for the initial consonants of the prefixes of 予 and 與 are based on other members of their phonetic series.

Similarly, he would have to reconstruct two roots for 'school', whereas I can reconstruct two la-roots or even just a single pre-OC √*lŋ:

And he would have to reconstruct three roots for 'lift', whereas I can reconstruct two la-roots or even just a single pre-OC √*lŋ:

How many other word families would be 'split' by Sagart's reconstruction?

I think it's unlikely that OC had at least three sets of *ɢ- and *l-roots with identical meanings. Hence I would avoid reconstructing uvulars in words which may have lateral-initial cognates.

Currently, I reconstruct uvulars in pure type A (emphatic) series with velars in Middle Chinese: e.g.,

后 *ɢoʔ 'sovereign'

侯 *ɢo 'target for shooting'; 'marquis' (latter same root as 后 or from 'shooter' [cf. Boodberg's translation 'scoparch' < scopus 'target']?; see this paper by Guillaume Jacques)

Mixed type A/B series with MC velars could consist of

- type A < original uvulars + type B < original uvulars shifted to velars via emphatic harmony with a nonemphatic presyllable

or

- type B < original velars + type A < original velars shifted to uvulars via emphatic harmony with an emphatic presyllable

External comparisons and statistical tendencies may help to identify the original type: e.g., the mixed (and predominantly type A) 古 series may have been originally type A uvular given

苦 OC *qha 'bitter' : Zhongu Tibetan qhɐ 'id.'

so type B 古 words must have had nonemphatic presyllables: e.g.,

居 OC *Cɯ-qa > *Cɯ-ka 'dwell'

(Emphatic harmony must have preceded the use of 居 to write the OC final particle *kə, now Md ji.)

08.11.28.23:00: Ə-NOTHER REASON TO DOUBT THE ZERO GRADE

I would like to write about something more positive for a change, but I also want to record my arguments against my own ideas before I forget them.

I proposed that the Old Chinese reflex of Proto-Sino-Tibetan (or pre-OC) zero grade *ŋ was *a, just as the Sanskrit reflex of the zero grade nasals *n and *m was a. The parallelism is illusory, as Skt a is a nonlow vowel that is a trace of an epenthetic schwa:

PIE *CN > *CəN > *Cə̃N > *Cə̃ > Skt Ca [Cə] or [Cɐ]

If Old Chinese were truly like Sanskrit, I would expect *ə instead of *a to be the reflex of pre-OC zero grade. Could these alternations (Sagart 1999: 61-62) be interpreted as *ə ~ zero grade alternations or as variant reflexes of an earlier zero grade *ŋ?

能 *Cʌ-nəŋ ~ *Cʌ-nə(ʔ) < ?*nŋ 'capable'

second form implied by Shijing rhyming as *-əʔ (Sagart 1999: 61; unable to confirm in Starostin 1989)

耳 *nəŋʔ ~ *nəʔ < ?*nŋʔ 'ear'

奶 *Cʌ-nəŋʔ ~ *Cʌ-nəʔ < ?*nŋʔ 'milk'

first form implied by Hakka nen

等 *Cʌ-təŋʔ ~ *Cʌ-təʔ < ?*tŋʔ 'wait'

齒 *ktəŋʔ ~ *ktəʔ < ?*tŋʔ 'tooth'

cluster *kt- to account for Amoy kh- and the *t-phonetic 止

These words have no semantic overlap, so they probably do not have a single suffix *-ŋ. Similarly, there is no semantic overlap between sets of Sanskrit verbs such as

Skt ma(n) 'think' and ha(n) 'kill'

Skt ga(m) 'go' and na(m) 'bend'

One cannot reconstruct an *-n suffix for the first pair or an *-m suffix for the second.

Sagart (1999: 62) accounted for the OC *-əŋʔ ~ *-əʔ alternations by proposing

an early shift of OC *-əŋʔ to *-əʔ, with dialectal preservation of *-əŋʔ.

(I have substituted my reconstructions for his.)

Here's a new hypothesis incorporating a change similar to Sagart's proposal:

In Proto-Sino-Tibetan or pre-Old Chinese, there was a distinction between *C-ŋ roots and *C-H-ŋ roots:

*H is an *a-coloring laryngeal (cf. Beekes' Proto-Indo-European *H₂ *[ʕ] > *a). *H by itself or in combinations with schwa (*Hə, *əH) became OC *a.

Like at least three other approximants, *H had a corresponding vowel:

All syllables with *H became OC emphatic syllables unless deemphasized by a preceding nonemphatic presyllable. (Note that I use underlining to avoid typing ˁ all over the place, and that ˁ is a miniature ʕ = *H.)

*Cɯ-CH(C)(ʔ)(s) > *Cɯ-Ca(C)(ʔ)(s)

Conversely, all syllables without *H became OC nonemphatic syllables unless emphasized by a preceding emphatic presyllable:

*Cɯ-Ca(C)(ʔ)(s) > *Cʌ-CH(C)(ʔ)(s)

Proposing *H just leads to more problems:

1. How can one determine whether a given OC *a is from earlier *a or *Hə? Maybe nonalternating OC *-aŋ is from original *a whereas alternating OC *-a ~ *-aŋ is from *-Hŋ ~ *-Həŋ, but what about OC *a before other codas and zero?

2. How many languages have numerous *Cʕ- clusters (if *H = *[ʕ])? The roots of all of these words would be reconstructed with such clusters in pre-OC: e.g.,

√*ɢʷʕŋ 'go'

I am suspicious of reconstructed languages with forms resembling random sequences of IPA characters.

3. How many languages have such exotic clusters in pronouns: e.g.,

√*ŋʕŋ 'I' (not √*ŋŋ as I had previously thought)

*ŋʕəŋ > 卬 OC *ŋaŋ 'I' (*ə-grade)

4. OC *a/*ə alternations require an infix *ʕ:

√nŋʔ 'thou'

*n(ə)ŋʔ > 乃 OC *Cʌ-nəŋʔ 'thy' (zero or *ə-grade)

*n-ʕ-ŋʔ > *naʔ > 汝 OC *Cɯ-naʔ 'thou' (*ʕ-infix, zero grade)

Does any language have such an infix? Googling "infix ʕ" results only in one hit. I could claim that this infix is a metathesized prefix

*ʕ-n-ŋʔ > *n-ʕ-ŋʔ > *naʔ > 汝 OC *Cɯ-naʔ 'thou'

but initial clusters like *ʕ-C- are still exotic.

What a mess. And I've only been looking at some of the Chinese data. I have no idea what to do with the *l-pronouns: e.g., are

余 *Cɯ-la < *-lHm 'I'

予 *Cɯ-la(ʔ) < *-lHm 'I'

台 *lə < ?*lm 'I'

zero grades of

朕 *r-ləm 'I'

(which is phonetic in *ləŋ-graphs, so are the above words zero grades of √lŋ?)

Last night, Guillaume Jacques reminded me that he had written a paper on Sino-Tibetan pronouns. I saw an earlier version in March and never got around to blogging about it. My apologies. Guillaume's paper discusses suppletive patterns which don't exist in Chinese but have to be reconstructed at the Proto-Sino-Tibetan level (or Proto-Tibeto-Burman, if one believes in that - neither Guillaume nor I do*):

I assume that the *q- of Qiang is archaic and I have left out vowels when they are uncertain.

I doubt there is any nonsemantic relationship between stem 1 and stem 2; *ŋ- and *q- or *n-ŋ and *k- are as unrelated as Sanskrit aham 'I' (< PIE √eg-) and maam (< PIE √me-) 'me'.

One could claim that stem 2 was a Proto-Tibeto-Burman innovation, but it is also possible that some languages (e.g, Qiang) retained stem 2 whereas others (Sinitic, Burmese, Tangut) independently lost it.

I suspect that the many Chinese pronouns are fossilized remnants of earlier declensional paradigms rivalling PIE in complexity: e.g., the different codas could be old case endings. Some ST languages still have paradigms (though one cannot assume they are entirely conservative). See Table 1 of Guillaume's paper for the many forms of Chang Naga pronouns.

*I do, however, find the term 'Tibeto-Burman' to be convenient shorthand for 'non-Sinitic Sino-Tibetan'. I use Matisoff's (2003) Proto-Tibeto-Burman as a source of generic composites of non-Sinitic ST forms: e.g., if I mention his PTB *naŋ 'thou', I am implying that two or more non-Sinitic ST languages have naŋ-like second person pronouns.

**Is there any non-Sinitic evidence for a *-ŋ in the Proto-Sino-Tibetan first person pronoun? Is Bodo-Garo *aŋ a metathesis of *ŋa to avoid initial *ŋ-?

08.11.27.22:51: AM I EXCEPTIONAL?

The earliest Old Chinese paradigm (late Shang period, c. 1300-1100 BC) is incomplete (tables below based on Sagart 1999: 142; all of Sagart's reconstructions in this post have been replaced by my own):

In the Western Zhou period (c. 1100-700 BC), 'I' was the odd man out in Old Chinese:

I don't know whether 爾 is truly an innovation. Was there a reason to avoid second person plurals in the Shang oracular inscriptions?

The second person pronouns are cognate, and the plural pronouns may share a suffix *-j. But the first person pronouns are noncognate in these early periods.

However, in the Eastern Zhou period, c. 700-255 BC), the first person pronouns become cognate:

What happened?

David Bradley's answer (paraphrased by Sagart [1999: 144])

a 1sg. pronoun *ŋa did exist in Old Chinese, but ... it was too ordinary or vulgar to occur in the inscriptions.

Sagart's counterargument

Here it should be recalled that the Shang oracular inscriptions were not made public, and that their language was not particularly literary of elevated in style; moreover ... in those Eastern Zhou inscriptions where 虍+魚 and 余 occur together, it does not seem that 余 is more literary, or less vulgar, than 虍+魚. Bradley's suggestion also seems to run against the usual pattern of evolution of personal pronouns: writing on the evolution of the Austronesian system of personal pronouns, R. Blust (1977: 11) observed that "in time all polite forms become ordinary, and new polite variants must be created". This entails the displacement of former ordinary forms and their eventual disappearance - the replacement of English thou by you comes to mind. The development envisioned by Bradley seems to go the wrong way: in it, an allegedly ordinary or vulgar pronoun *ŋa competes with, and eventually displaced, an allegedly polite form *Cɯ-la.

Sagart's answer

Actually, the replacement of *Cɯ-la by *ŋa does not require us to suppose that a first-person pronoun *ŋa already existed in Shang times [though it was not written]. The emergence of *ŋa in Chinese of around 700 BCE, and the disappearance of *Cɯ-la, are easily explained in terms of analogy ... the replacement of the Western Zhou 1sg. pronoun *Cɯ-la by*ŋa makes the Eastern Zhou pronoun system almost entirely symmetrical: we are in the presence of a typical instance of proportional analogy.

I agreed with Sagart's answer for eight years, even though it does not account for the absence of a glottal stop in 虍+魚/吾*ŋa 'I' corresponding to the glottal stop in 汝 *Cɯ-naʔ 'thou'.

My answer as of Monday morning (but not anymore)

The Proto-Sino-Tibetan first person pronoun root was √*ŋŋ. It would be difficult to distinguish between zero grade *ŋŋ 'I' and *nŋʔ 'thou', so the former was displaced by 余 *Cɯ-la, whereas the latter developed regularly into the root of 汝 *Cɯ-naʔ in early Old Chinese. (Perhaps both had the same OC prefix.) Early Old Chinese also had an *a-grade first person pronoun 卬 *ŋaŋ 'I' absent from Sagart's tables.

In Tibeto-Burman, on the other hand, zero grade *ŋŋ 'I' developed regularly into *ŋa, and the second person pronoun was *a-grade: *naŋ. (Could Tibeto-Burman forms like Jingpho nā [Matisoff 2003: 264] reflect zero-grade *nŋʔ?)

In Early Old Chinese, 我 'we' was *ŋaj from *ŋa, the zero grade of √*ŋŋ (with automatic emphasis conditioned by a low vowel), plus the plural suffix *-j also in 爾 *Cɯ-najʔ 'you' < *Cɯ-naʔ + *-j (with metathesis).

Later, 我 *ŋaj 'we' became *ŋajʔ by analogy with 爾 *Cɯ-najʔ 'you'.

(New twist on Thursday: I just remembered that Meixian Hakka 我 ŋɔ 'I' has a yinping tone possibly implying earlier *s-ŋaj without a glottal stop. Could Hakka be the only branch of Sinitic that lacks an innovative glottal stop in 我, or is the tone by analogy with ping tone 吾 'I'? The function of the prefix needed to account for the yin [< *voiceless onset] category is unknown.)

(Suzhou and Shanghai have yangqu tone for 'I' possibly implying earlier *ŋajʔ-s or *ŋaj-s. The function of the suffix [or tone shift] is unknown.)

I remain troubled by the zero grade hypothesis. I have already given several reasons in the last two posts. Here's another: if zero grade was a Proto-Sino-Tibetan phenomenon, is there any evidence for it in Tibeto-Burman? Should Jingpho pairs like

nā ~ nāŋ 'you' (Matisoff 2003: 264)

gùmrà ~ gùmràŋ 'horse' (Matisoff 2003: 177)

cf. Old Chinese 馬 *mraʔ 'horse' < zero grade *mrŋʔ?

and

TT5233 rieʳ R79 1.74 'horse' < *Ci-ra < zero grade *rŋ?

as well as many TB -ŋ 'horse' words

be explained via suffixation

'you': *na(-ŋ)

'horse': *mra(-ŋ)

'you': *nŋ ~ *naŋ

'horse': *mrŋ ~ *mraŋ

Are there any other potential -a ~ -aŋ alternations in Tibeto-Burman?

08.11.26.3:06: YOU AND I CAN'T SOUND ALIKE, OR I MET RESISTANCE

For a couple of days, I've been bothered by the improbability of the sound change

*ŋŋ > *ŋa

required by my zero grade hypothesis. I've also been thinking that *ŋŋ (initial plus syllabic nasal) would be difficult to distinguish from a syllabic nasal *ŋ. How could anyone hear something like [ŋ] in rapid speech and split it into [ŋ ... ŋ] in non-zero grade forms? Moreover, the zero grade of the root √*nŋʔ that I reconstruct for 'thou' is almost indistinguishable from √*ŋŋ 'I'/'meet':

Languages can have soundalike pronouns: e.g., Korean

na 'I' : nE-ga (< *na-i-ka) 'I' (nominative)

nO 'thou' : ne-ga (< *nə-i-ka) 'thou' (nominative)

11.27.22:55: ne-ga can also be ni-ga to differentiate it from nE-ga which many pronounce as ne-ga due to the merger of E with e

11.27.23:05: One could view these Korean pronouns in terms of a chain shift:

nE-ga > ne-ga > ni-ga

but I've never heard of a pair of soundalikes like *ŋŋ and *nŋʔ.

Tonight I realized that Sanskrit has no roots like √n-n and √m-m. I presume no such roots were in Proto-Indo-European. If a Sanskrit root has two nasals, they are heterorganic:

√mn 'think':

man-iṣya-te 'will think' (a-grade)

ma-ta- 'thought' < *mn̩- (zero grade)

√nm 'bend':

nam-iṣya-ti 'will bend' (a-grade)

na-ta- 'bent' < *nm̩- (zero grade)

Yet according to my zero grade hypothesis, pre-Old Chinese had not just one but two homorganic nasal roots: √*ŋŋ 'I' and √*ŋŋ 'meet'. Although it is possible that pre-OC was a sui generis language, I prefer reconstructions with parallels to known languages.

(It just occurred to me that I don't remember Pulleyblank mentioning the zero grade in his writings about Sino-Tibetan and Indo-European. He's believed in a genetic connection between the two for decades. I wonder if he considers IE zero grade to be an innovation absent from ST.)

08.11.25.3:12: I OPPOSE TEETH

Three members of the Old Chinese 'meet' word family were written with the 'motion' radical 辶 plus three different phonetics:

迎 *Cɯ-ŋaŋ 'go to meet' (with phonetic 卬 *ŋaŋ 'I')

逆 *Cɯ-ŋak 'go against' (with homophonous phonetic 屰 'id.')

迓 *ŋa-s 'meet' (with phonetic 牙 ?*r-ŋa 'tooth' [see here])

Should the root of this trio be reconstructed as *ŋa

迎 *Cɯ-ŋa-ŋ

逆 *Cɯ-ŋa-k

迓 *ŋa-s

迎 *Cɯ-ŋaŋ < *Cɯ-ŋaŋ (*a-grade)

逆 *Cɯ-ŋak < *Cɯ-ŋŋ-C (zero grade plus assimilating stop suffix)

or *Cɯ-ŋaŋ-C (*a-grade plus assimilating stop suffix)

迓 *ŋa-s < *ŋŋ-s (zero grade plus fricative suffix)

If the root ended in a vowel, I would expect derivatives ending in nonvelar codas other than *-s. Schuessler (2007: 551) proposed several such derivatives, but I think they belong to separate word families:

*ŋa 'riverbank':

岸 OC *ŋa-n-s 'riverbank'

厂 OC *hŋa-n-s 'riverbank' < ?*s-ŋ-

滸 OC *hŋa-ʔ 'riverbank' < ?*s-ŋ-

Written Tibetan dngo 'shore, bank' (unexpected vowel)

*ŋar 'face, front side':

顏 OC *ŋran < ?*ŋar 'face' (see Schuessler 2007: 86)

Written Tibetan ngar 'front side'

OC *ŋre < *N-kre 'shore'

崖涯 with stop-initial phonetic 圭 *kwe

I'm not convinced by my attempts to link their meanings with 'meet':

'riverbank', 'shore' < 'where water meets land'?

'face, front side' < 'what is met'?

Schuessler (2007: 551) could not find any Tibeto-Burman 'meet' words ending in -k. He concluded that OC "open vowel and final *ŋ forms are inherited from ST [Sino-Tibetan] ... The [OC] forms in final -k are CH innovations." Did Proto-Sino-Tibetan had an *-a ~ *-aŋ alternation derived from pre-PST: *-ŋ ~ *-aŋ?.

Is there any evidence for other zero grade syllabic sonorants in (pre-)PST? I don't know of any zero-grade ~ *a-grade alternations like

*-a < *-r ~ *-ar

*-i < *-n ~ *-an

*-u < *-m ~ *-am

*-u < *-l ~ *-al

I am hesistant to regard

語 *Cɯ-ŋaʔ 'speech' ~ 言 *Cɯ-ŋan 'speak'

as being from

*Cɯ-ŋn-ʔ (zero grade) ~ *cɯ-ŋan (*a-grade)

because it probably involves a suffix*-n that can also appear after vowels other than *a (see Schuessler 2007: 74-76 for examples).

There is no reason for zero grade to only exist before *-ŋ. Therefore the zero grade hypothesis is improbable. Nonetheless, the frequency of Old Chinese *-a ~ *-aŋ alternations compared to other *-V ~ *-VN alternations still requires an explanation.

08.11.23.23:59: 于ING TO ZERO?

When I first studied Sanskrit, I didn't even think about alternations like

han-iṣya-ti 'will kill' <> ha-ta- 'was killed'

gam-iṣya-ti 'will go' <> ga-ta- 'gone'

I just took it for granted that root-final -n and -m sometimes disappeared. Later, I learned that the nasalless forms once contained nasals - and no vowels:

ha-ta- < *ghn̩-to-

ga-ta- < *gwm̩-to-

Syllabic *n and *m had merged into a. Reconstructing 'zero grade' (i.e., vowelless) past participles restores the parallelism between the above forms and the corresponding forms of mṛ 'die':

I wonder if Old Chinese *-aŋ ~ *-a alternations also originally involved a 'zero grade'. The following list is based on Gong (1995: 58) but the reconstructions are mine:

It is curious that all of these alternations involve nonemphatic words with the exception of 卬 ~ 吾 which is probably not a true alternation. 吾 is newer than 卬 and must postdate the shift of syllabic *ŋ to *a which would have occurred before the invention of sinography. (There is no graphic or rhyming evidence for syllabic *ŋ, so such a sound would have to predate literacy.) The *-a of 吾 'I' may be by analogy with the *-a of 汝 *Cɯ-naʔ 'thou' (Sagart 1999: 144):

If the singular of 爾 *Cɯ-najʔ 'you' is 汝 *Cɯ-naʔ 'thou',

then the singular of 我 *ŋajʔ 'we' is 吾 *ŋa 'I' (but why isn't it *ŋaʔ?)

Some of the Old Chinese *a-forms could postdate the shift of syllabic *ŋ to *a if they were created by analogy with older zero grade forms.

Sagart (1999: 134-135) accounted for these alternations with a suffix *-ŋ, but most of the above words do not have cognates with other codas: e.g., one might expect a root ending in *-a to have derivatives ending in *-a-n, *-a-t, etc. as well as *-aŋ.

我 *ŋajʔ 'we' is presumably cognate to 卬 *ŋaŋ 'I'. Sagart (1999: 135): derived *ŋaŋ from *ŋajʔ-ŋ, whereas I wonder if *ŋajʔ might be from a zero grade *ŋŋ plus suffixes.

11.24.0:15: I presume 我 *ŋajʔ 'we' (< ?*ŋŋjʔ) has the same plural (?) suffix *-j as 爾 *Cɯ-najʔ 'you' which may be derived from 汝 *Cɯ-naʔ 'thou'. The final glottal stop of 我 *ŋajʔ is absent from 卬 *ŋaŋ 'I' and may have been added by analogy with 爾 *Cɯ-najʔ.