Digital resources in the Social Sciences and Humanities OpenEdition Our platforms OpenEdition Books OpenEdition Journals Hypotheses Calenda Libraries OpenEdition Freemium Follow us
Image Image

Posts

Posted on

PAN *water’: *Zanum and *nanum

(view tree) (view list of innovations, sorted by node)

The PAN word for ‘water’ was *ZaNum [ɗalum]. This is phonemically more specific than Blust’s *daNum, since  Blust’s *d conflates three distinct PAN phonemes (Ross 2015). I am using Ross’s symbols.  In Sagart (2025) I proposed that Ross’s *Z was implosive: [ɗ]. Blust’s gloss ‘fresh water’ may be too restrictive as there is no competing form for salt water at PAN level.

A verbal meaning may be supposed at a post-PAN level, although it is only attested in Amis radom, mi-radom ‘carry, draw or pour out water’ (Wu Ming-yi 2013).

Out of the verb, a noun was derived by the <in> infix: *Z<in>aNum [ɗinalum] ‘water carried home’, hence ‘drinking water’.

The new trisyllable was irregularly shortened to a disyllable through left-pruning, being too long for its frequency (cf. Manczak’s ‘Irregular Sound Changes due to Frequency, Manczak 1969, 1977, 1978, 1980). This resulted in *naNum [nalum].

Coarticulation from the initial and final nasals gave rise to variant pronunciations with nasalized lateral [nal̃um]; that variant, reinterpreted phonemically as /nanum/, was preferred to inherited /nalum/ as it allowed a more robust distinction with *ZaNum /ɗalum/.

The phonetic sequence leading to *nanum was completed at the Eastern Walu-Siwaish node, of which *nanum ‘(…), drinking water’ is a defining innovation. *nanum competes with *ZaNum in languages of the Formosan east coast, Kra-Dai and Malayo-Polynesian.

Doublets for *ZaNum and *nanum exist in Formosan, MP and KD languages: https://stan.hypotheses.org/files/2026/01/Screenshot-2026-01-25-at-14.54.45.png

References

Mańczak, W. 1969. Le developpement phonétique des langues romanes et la fréquence, Kraków: Uniwersytet Jagielloński.

Mańczak, W. 1977. Slowiańzska fonetyka hisoryczna a frekwencja [Slavic historical phonetics and frequency], Kraków: Uniwersytet Jagielloński.

Mańczak, W. 1978. Irregular Sound Change due to Frequency in German, in Recent Developments in Historical Phonology, pp. 309-19, J. Fisiak (ed.), The Hague: Mouton.

Mańczak, W. 1980. Irregular Sound Change due to Frequency in Latin. Language Sciences, volume 2, number 1, 62-68.

Ross, M 2015, ‘Some Proto Austronesian coronals reexamined’, in Elizabeth Zeitoun, Stacy F. Teng & Joy J. Wu (ed.), New Advances in Formosan Linguistics, Asia-Pacific Linguistics, Canberra, Australia, pp. 1-38.

Sagart, Laurent. “The PAN Voiced Stops and Affricates” Oceanic Linguistics. Published electronically May 3, 2025. https://dx.doi.org/10.1353/ol.2025.a960256.

Wu Mingyi (Namoh Rata). 2013. O Pidafo’an to Sowal Misanopangcah [dictionary of the Amis language]. Taipei: Nan t’ien.

Posted on

OC voiceless sonorants: a second look (with Bill Baxter)

Since the publication of Old Chinese: A New Reconstruction (Baxter and Sagart 2014, hereafter ‘OCNR’), solid proposals to change aspects of the OC reconstruction in that book, whether at the level of individual words or at a more systematic level, have appeared. On our part, we have accumulated new ideas in a database that we maintain. The time is ripe for an updated version of the OCNR system. Our plan is to work toward that goal, hopefully in collaboration with interested colleagues; and with luck, to complete it. At any rate it is our hope that younger scholars will continue improving the OCNR system after us. At this point it becomes essential for us to begin making public some of the lines of reconstruction which, we feel, should be part of a revised system. One of these relates to voiceless sonorants.

In what follows, we use lenticular brackets 【】 to indicate OC reconstructions which we feel should replace the corresponding  OCNR forms. By default, reconstructions not in lenticular brackets follow OCNR.

Cross-linguistically, voiceless sonorants usually result from the devoicing of sonorants by adjacent voiceless sounds, in whatever order. That is probably also the origin of the OCNR voiceless sonorants. Judging from word-family connections like 墨 MC mok > mò ‘ink’ and 黑 MC xok > hēi ‘black’, which implies a root *mˤək, the order of segments was: voiceless obstruent + sonorant. When did the sonorants devoice? According to Mei (2012), after Old Chinese; and the devoicing consonant was *s-. According to us (Sagart and Baxter 2012, Baxter and Sagart 2014), they were already devoiced in Old Chinese. We made no claims concerning the identity of the devoicing segment(s), beyond that it was voiceless.

Consider chǒu 丑 ‘2d earthly branch’, MC trhjuwX, OCNR *[n̥]ruʔ. This word provides evidence that a voiceless stop preinitial, namely *p., still present around 200 BCE in at least some forms of Chinese, is responsible for the voiceless nasal in OCNR. The Chinese cyclical signs were borrowed by the Tai languages as year names in the early times of Chinese-Tai contact, beginning in 203 BCE when Zhào Tuó 趙佗, at the head of a Chinese army, established the Nányuè 南越 state near present-day Guǎngzhōu. Like the Chinese numerals, borrowed at the same time, they are phonologically archaic. Tai words corresponding to 丑 *[n̥]ruʔ > trhjuwX > chǒu ‘2d earthly branch’ include: Ahom plāu, Lü pau³, Dioi piaou³, Pu-yi piu³ (Norman 1985:86): only the vowel and tone seem to match. How to explain the p- ? it must have been there in the donor language, even though it is attested nowhere, as far as we know, in Chinese. Reconstructing 丑, then, as OC 【*p.nruʔ】, it is plausible that early Tai speakers rendered this as /pn-/ or /pr-/, which then evolved to the cited modern forms. In Chinese, the voiceless preinitial devoiced the nasal and fell: 【*p.nruʔ】> mid-OC *p.n̥ruʔ > late OC *n̥ruʔ. We could have reconstructed OC *p.n̥ruʔ in OCNR, but since plain and voiceless sonorants do not contrast after voiceless preinitials, it is economical to explain voicelessness in the sonorant as spread from the preinitial. On this account, voiceless sonorants only existed in mid- or late OC, originating in OC tight clusters of a voiceless obstruent and and a sonorant, without an intervening vowel. Additional examples follow:

血 OCNR *m̥ˤik > xwet > xuè ‘blood’. Xiaolian Bai shuã 44 (note the vowel nasalization, pointing to a lost nasal; tone 44 indicates an earlier stop ending) argues the devoicing segment was *s-. This is now 【*s.mˤik】with *s.m- > m̥- > xw-.

捝 OCNR *l̥ˤot > thwat > tuō ‘peel off, take off’. This is now【*k-lˤot】with *k.l- > l̥- > th– . Proto-Kra (Ostapirat 2000) *klut ‘take off’ for the velar preinitial.

鐵 OCNR *l̥ˤik > thet > tiě ‘iron’ and 梯 OCNR *l̥ˤ[ə]j > thej > tī ‘stairs’, Mulam kɣet and kɣø 3. Both words are now reconstructed with OC【*k.l-】, with *k.l- > l̥- > th-.

攝 OCNR *kə.n̥ep > syep > shè ‘catch, gather up’ where the velar preinitial is indicated by Lakkia khjɛ̃:p < *k[n]ep and Ruc kăɲă’p ‘close’ (the eyes). This now becomes【*k.nep】with *k.n- > n̥ > MC sy-.

Gong (2017) has identified more examples in a mediaeval Vietnamese text which uses a restricted set of Nôm characters to write preinitials in inherited mon-Khmer words. The same characters write preinitials in Chinese loanwords. Thus mediaeval Vietnamese 巴拭 p-ɕứk Gong’s reconstruction) writes 拭 ‘to wipe’, OCNR *l̥ək > syik > shì. This must now be OC【*p.lək】> mid OC *p.l̥ək > late OC p.ɕək > MC syik. The Vietnamese form reflects the late OC stage. It shows the the l̥ > sy change completed and the presyllable still in place. Similarly, Gong reasonably identifies the Nôm strings 可耶, 可賒, 可車, mediaeval Vietnamese *k-ɕa ‘far’ as rendering Chinese 賒 ‘long-term, distant, OCNR *l̥A > syae > shē’ as the loanword’s source. This word must now be OC【*k.la】, borrowed at the k-ɕa stage.

In the numerous cases where the devoicing preinitial is unknown, we write a generic voiceless consonant ‘C̥’, e.g.:

海 xojX < OCNR *m̥ˤəʔ, now 【*C̥.mˤəʔ】

手 syuwX < OCNR *n̥uʔ, now 【*C̥.nuʔ】

埶 syejH < OCNR *ŋ̊et-s, now 【*C̥.ŋet-s】

申 syin < OCNR *l̥i[n], now 【*C̥.li[n]】

爍 syak < OCNR *r̥ewk, now 【*C̥.rewk】

etc.

In all examples so far, the voiceless preinitial falls after devoicing a sonorant, and is invisible as a segment in MC. But there are also a number of forms where the MC onset appears to go back to a sonorant root initial, devoiced by a preceding voiceless segment which is still visible, or has left a trace, in MC. In the OCNR framework such words are handled by reconstructing an OC voiceless preinitial followed by a voiceless sonorant, e.g.:

千 OCNR *s.n̥ˤi[ŋ] > tshen > qiān ‘thousand’ (the phonetic is 人*niŋ);

闃 OCNR *[k-m̥]ˤik > khwek > qù

川 OCNR *t.l̥u[n] > tsyhwen > chuān ‘stream, river’.

We assume that at the moment when devoicing preinitials normally fell, in 千, 闃 and 川, the inner preinitial (tightly attached) was itself preceded by an unknown, loosely attached outer preinitial. The inner preinitial syllabified with the preceding vowel, which allowed it to remain longer. Rather than using a cumbersome *(C)V. notation, we indicate the presence of an outer preinitial with a hyphen preceding the inner preinitial. Thus we write 【*-s-nˤiŋ】for 千, MC tshen, against just【*s.niŋ 】 for 身, MC syin. Likewise 闃 OCNR *[k-m̥]ˤik is now 【*-k-mˤik】, and 川, OCNR *t.l̥u[n] is 【*-t-lu[r]】.

Now OCNR already had forms with a single voiceless preinitial consonant directly followed by a sonorant, without the MC initial being devoiced. These reconstructions must now be modified. We need to suppose that the preinitial in them was of the loosely attached kind, with a vowel separating the voiceless preinitial and the root-initial sonorant, preventing devoicing. In another change from OCNR-style reconstructions, we write that vowel as a generic ‘V’ because OC preinitials may have contained different vowels.

袂 OCNR *k.mˤet > kwet > mèi ‘sleeve’, now 【*kV-mˤ[ek]】. Also read mjiejH.

舞 OCNR *k.m(r)aʔ > mjuX > wǔ ‘dance (v.)’, now【*kV.m(r)aʔ】. Rục kumu’a.

肉 OCNR *k.nuk > nyuwk > ròu ‘meat, fesh’, now【*kV.nuk】. Ferlus Proto-Pong *kɲuk⁷ ‘chair, viande, flesh’

多 OCNR *[t.l]ˤaj > ta > duō ‘many’, now 【*tV-lˤaj】. Phonetic-series contact to 移 OCNR *laj > ye > yí ‘move (v.)’ and Proto-Tai *hlaj A ‘much, many’ (Li 1977). Note that Li reconstructed *hl- in this word on account of high-series tones in the modern reflexes. No voiceless initials are found in the daughter languages: Li’s PT form is consistent with our【*tV-lˤaj】.

質 OCNR *t-lit > tsyit > zhì ‘substance, solid part’, now【*tV-li[t]】. Word-family contact to 實 OCNR *mə.li[t] > zyit > sh. ‘fruit; full’

絞 OCNR *k-rˤ[i]wʔ > kaewX > jiǎo ‘twist; strangle’ now【*kV-rˤ[i]wʔ】. Word-family contact to 摎 *[r]iw > ljuw > liú ‘tie around, strangle’

萬 OCNR *C.ma[n]-s > mjonH > wàn ‘10,000’ now 【*tV.ma[n]-s】. Manchu tumen, Old Turkic tümen, Tokharian A tmāṃ ‘10,000’.

Preservation or not of the preinitial consonant in MC is unpredictable, as CV preinitials behave erratically across East Asia. This has been repeatedly observed in Chamic (Thurgood), Vietic (Ferlus), Kra (Li Jinfang). We are dealing with a lexical, phonologically non-regular change. This results in doublets, for instance 袂 ‘sleeve’, MC kwet and mjiejH above, both derivable from【*kV-m-】.

Here again, when we lack evidence on the identity of the preinitial consonant, we write a generic ‘C̥V’.

The present model goes back to a passage in Sagart and Baxter (2012:44-45) where we responded to criticism by Mei (2010). Mei had argued that if we reconstruct voiceless sonorants for words like 烕 *m̥et, and *s- plus sonorants in words like 戌 *s-mit, we are unable to connect the former with Tibeto-Burman words reconstructed with *s- plus sonorants. As we pointed out, one can easily imagine scenarios where our OC voiceless sonorants are cognate to the *s-clusters of Tibeto-Burman, illustrating one such with the following table:

Hypothetical pre-OC Old Chinese Middle Chinese Mei’s reconstruction
*s.m-, *s.n-, *s.ŋ- *m̥-, *n̥-, *ŋ̊- x, th, x *sm-, *sn-, *sŋ-
*sə.m-, *sə.n-, *sə.ŋ- *s.m-, *s.n-, *s.ŋ- s ?

Origins of OC voiceless nasal: a hypothetical scheme. Reproduced from Sagart and Baxter (2012:44).

Under this scheme, there is no difficulty in connecting Tibeto-Burman forms like WT snying ‘heart, mind’, with 身 OCNR *n̥i[ŋ] > syin > shēn ‘body; self’ (now 【*[s]-ni[ŋ]】) . We can also connect Tibeto-Burman s-nasals with our OC *s-nasals, e.g. Written Burmese hnaŋ A, possibly out of PTB *s-naŋ ‘to drive, drive along, drive away’, opposite 襄 OCNR *s-naŋ > sjang > xiāng ‘remove’ (now 【sV-naŋ】).

As to why our OCNR reconstruction does not follow the model in our table, we wrote that we knew of no evidence pointing in the direction of voiceless preinitials in forms we reconstruct with OC voiceless sonorants. That is no more the case: the scheme in the above table now becomes the best explanation of the facts. ‘Hypothetical pre-OC’ in the table becomes ‘Old Chinese’, and ‘Old Chinese’ becomes ‘Late Old Chinese’’, with the understanding that any voiceless preinitial, not just *s, as long as it is tightly attached, devoices any sonorant that follows it.

References

Baxter, William H. & Laurent Sagart. 2014. Old Chinese: a new reconstruction. New York: Oxford University Press.

Li, Fang Kuei. 1977. A handbook of comparative Tai (Oceanic Linguistics Special Publication 15). Honolulu: University of Hawai’i Press.

Mei Tsu-lin. 2010. The causative *s- and nominalizing *-s in Old Chinese and related matters in Proto-Sino-Tibetan. Paper presented at the International Symposium on Sino-Tibetan Comparative Studies in the 21st Century, Nankang, June 24-25, 2010.

Norman, Jerry. 1985. A note on the origin of the Chinese duodenary cycle. In Graham Thurgood, James A. Matisoff & David Bradley (eds.), Linguistics of the Sino-Tibetan area: the state of the art: papers presented to Paul K. Benedict for his 71st birthday, 85–89. Canberra, A.C.T., Australia: Dept. of Linguistics, Research School of Pacific Studies, the Australian National University.

Ostapirat, Weera. 2000. Proto-Kra. Linguistics of the Tibeto-Burman Area 23(1). 1–251.

Sagart, Laurent & William H. Baxter. 2012. Reconstructing the *s- prefix in Old Chinese. Language and Linguistics 13(1). 29–59.

Posted on

The *i vs. *ə alternation in the Sino-Tibetan word for ‘10’

Some generalizations for a start. Those Sino-Tibetan (ST) words for ‘10’ that end in a labial consonant have the widest distribution and almost certainly reflect the main PST etymon for ‘ten’. They come in two flavors: with vowels *i and *ə. The i-set is limited to Tibeto-Burman (TB). Its members either end in a stop: -ip, -ep, -iap: WB kyip, Limbu gip, Mikir kep, Geman Deng kiɑp/kyep, etc. or—if the stop has been lost— in the vowel preceding it: for instance in some rGyalrongic languages: Japhug sqi, Ergong (Daofu); (Huang and Dai 1992) zʁa ‘ten’, sqha (‘ten’ in ‘20’, ‘30’ etc., from earlier zʁe, sqhe, or zʁia, sqhia). The STEDT website reconstructs *gip ‘ten’, which is fine except that Japhug sqi and Ergong zʁa indicate a uvular initial. Under Matisoff, the STEDT ignores the ST distinction between velars and uvulars. WB kyip, on the other hand, seems to argue against a uvular, since PST uvulars normally go to zero/ʔ (depending on the transcription being used) in Burmese (Peiros and Starostin 1995). Here we probably see the original uvular anteriorized to a velar by the preceding s- in rGyalrongic, a closely related group according to Sagart et al. (2019). This points to a PTB *sɢip.

The STEDT does not give a reconstruction for the ə-vowel set. TB merges *ə into *a, thus the TB forms in this set usually have a low vowel. They include Proto-Tani *čam ‘tens’ (Sun 1993) as in Adi Bokar ʧam– ’10’ in numbers ‘20’ and above. The situation in rGyalrongic is more complex. As already mentioned, Japhug has sqi for ‘10’. In Jacques’s synchronic analysis (2021:236), ‘10’ as a formative in teens is is /sqa/ . Specifically in 11, 12, 13 and 16 it is immediately followed by a labial consonant before the unit number: 11 sqa-p-tɯɣ, 12 sqa-m-nɯz, 13 sqa-f-sum, 16 sqa-p-rɤɣ. The labial element is not inserted in 14 and 15, as the unit number already begins in a labial: sqa-βde, sqa-mŋu. No labial element is inserted in 17-19: sqa-ɕnɯz, sqa-rcat, sqa-nŋɯt. It is of course the lack of an inserted labial in 17-19 which motivates treating -p-, -m-, -f- as insertions: historically they are vestiges of an earlier *-p coda on ‘10’ (pre-Japhug *sqap) assimilated on the following numeral’s initial. Ergong (Huang and Dai 1992) is parallel to Japhug: along with zʁa and sqha, already mentioned for ‘ten’ and its exact multiples, one has ʁɑ in teens (notice the back vowel). The -m endings in Proto-Tani *čam ’tens’ probably also had their *-p ending assimilated on a following nasal: for an unknown reason the *-am forms displaced the *-ap forms. If we wanted to include Proto-Kuki-Chin *soom ‘ten’ (VanBik 2009) in the ə-vowel set, we would need a change ɢ > w, so that *sɢaap/m > swaam > soom. I am not aware of any evidence for that change (noyt that I would necessarily), so the reason for the vowel change may well lie elsewhere.

rGyalrongic holds clues to a solution to the two-flavor problem. Japhug and Daofu have a front vowel when ‘ten’ occurs at the end of a numeral expression—i.e. in ‘10’ and its multiples—vs. a low or back vowel when ‘10’ is non-final, being followed by a unit numeral—in teens. If PST numeral expressions were stressed on the last morpheme and non-final morphemes were unstressed, non-final ‘ten’ would have a more central vowel than word-final ‘10’ phonetically. This could eventually phonologize so that one would get a phonemic contrast between /sɢip/ and /sɢəp/. Only the latter’s ending would have the opportunity to assimilate on a following numeral’s onset—the former was never followed by a numeral. This will explain why the *p ending in the *i-set either remains or falls, but never assimilates, and why only the *ə-set can end in -m.

Tani has lost the i-vowel variant to an innovated form Proto-Tani *rjɯŋ ‘ten’. In all other contexts the ə-variant *čam ‘tens’ has prevailed, even in exact multiples of ten because the Tani order for these numerals is 10-multiplicator, e.g. Adi-Bokar tʃɑmɲi 10-2 ‘twenty’, tʃɑmum 10-3 ‘thirty’, tʃɑmpi 10-3 ‘forty’, tʃɑmŋo 10-5 ‘fifty’. Here we understand why.

Let us now turn to Chinese. The Baxter-Sagart OC reconstruction is *t.[g]əp > MC dzyip ‘10’. This is self-evidently part of the ə-vowel set.  *[g] is shorthand for ‘either *g or *ɢ’. Although after non-pharyngealized onsets OC *-ip and *-əp are indistinguishable—both MC -ip—we are confident that in Old Chinese ‘10’ had OC *-əp because c. 1300-1200 BCE the Oracle Bone inscription graphs for shí 十 ‘10’ and for zhēn 箴/鍼/針 ‘needle’ are identical: a vertical stroke. In some bronze inscriptions one end of the stroke is blunt and the other is sharp: clearly a drawing of a needle. Both the loan to Vietnamese găm ‘bamboo or metal pin’ and the WB cognate word ap ‘needle’ are incompatible with *-ip/im. Baxter and Sagart (2014) reconstructed ‘needle’ 箴鍼針 as *t.[k]əm, where [k] is shorthand for *k and *q. Proto-Tai *qem ‘needle’ (Pittayaporn 2009) is a Chinese loanword whose *q- could reflect OC pharyngealized *kˤ-, but not plain *k-. This disambiguates ‘needle’ to OC *t.qəm. Thus ‘10’ and ‘needle’ were phonetically close: *t.ɢəp vs. *t.qəm. Normally the difference in endings would have stood in the way of equivalence in the script. Either ‘needle’ had a doublet with -p (as in WB ap, WT khab, Japhug taqaβ), or ‘10’ had a doublet *t.ɢəm through assimilation on a numeral beginning in a nasal, as in rGyalrongic.

It is as yet unclear if the TB and Chinese preinitials: *s- vs. *t., are relatable. So for PST I suggest a doublet for ‘10’: expression-final *[t,s]V.ɢip, non-final *[t,s]V.ɢəp. Chinese has generalized the non-final form across the entire paradigm. As we saw, TB maintains the two and the context in which the PST vowel distinction first arose is only displayed by rGyalrongic.

References

Baxter, William H. and Laurent Sagart. 2014. Old Chinese: a new reconstruction. New York: Oxford University Press.

Jacques, Guillaume. 2021. A grammar of Japhug (Comprehensive Grammar Library 1). Berlin: Language Science Press. http://langsci-press.org/catalog/book/295

Huang Bufan and Dai Qingxia, eds. 1992. Zangmianyuzu yuyan cihui 《藏緬語族語言詞匯》[A Tibeto-Burman Lexicon]. Beijing: Central Institute of Minorities. Accessed via STEDT database <http://stedt.berkeley.edu/search/> on 2024-10-31.

Peiros, I., and S. Starostin. 1995. A comparative vocabulary of five Sino-Tibetan languages. 5 fascicles. Melbourne: Department of Linguistics and applied Linguistics, University of Melbourne.

Pittayaporn, Pittayawat. 2009. The phonology of proto-Tai. Unpublished Cornell U. dissertation.Laurent Sagart, Guillaume Jacques, Yunfan Lai, Robin J. Ryder, Valentin Thouzeau, Simon J. Greenhill, Johann-Mattis List. 2019. Dated language phylogenies shed light on the ancestry of Sino-Tibetan. Proceedings of the National Academy of Sciences May 2019, 116 (21) 10317-10322.

Sun, Jackson Tianshin. 1993. A historical-comparative study of the Tani (Mirish) branch in Tibeto-Burman. Ph.D. Dissertation, University of California, Berkeley.

VanBik, Kenneth. 2009. A reconstructed ancestor of the Kuki-Chin languages. STEDT monograph 8. Berkeley, University of California.

Web sites

STEDT database <http://stedt.berkeley.edu/search/> on 2024-10-31.

Numeral systems of the world’s languages. https://lingweb.eva.mpg.de/channumerals/

Posted on

A proto-Kra-Dai preinitial

[Added July 12, 2024. This post is in error.  The morpheme in words beginning in Baoding tsɯ²-, Heitu kɯ2– , Tongshi ʔɯ3  is a general nominal classifier, glossed as ‘一(一个)’ in Ouyang and Zheng (1983:510). The apparent correspondence with Buyang qa- must be coincidental. Oops !]

The three principal subgroups of Kra-Dai are Kra (a.k.a Ge-Yang), Hlai (Hainan Island) and  Kam-Tai. While many Kra-Dai (or Tai-Kadai) languages are largely monosyllabic, in some languages main syllables are preceded by CV minor syllables (‘preinitials’) which correspond to the initial syllables of Austronesian words, as first noted by Li Jinfang for the Kra language Buyang, where an unusually large number of these forms can be found. A list of these, with their Austronesian cognates, was given in Sagart (2005). Guillaume Jacques (2017) has studied the correspondences of the Buyang (Langjia) preinitial qa0– in the Kam-Tai branch.

A correspondence for Buyang preinitial qa0– has now been observed in Hlai dialects: tsɯ²- in Baoding and kɯ2– in Heitu. Although not shown here, the corresponding words in Tongshi, also a Hlai dialect, have preinitial ʔɯ3. Buyang Langjia data are from Li Jinfang (1999); Hlai data from Ouyang and Zheng (1983). The first three words originate in Austronesian prototypes beginning in *q-.

source

Buyang

Baoding

Heitu

shoulder

PAN qabaRaX

qa⁰ʔba¹¹

tsɯ²va²

kɯ²va²

mortar

PAN qaS1elu

—

tsɯ²rau¹

kɯ²rau³¹

nail

PAN qurip

qa⁰lip¹¹

tsɯ²li:p7

—

grass ash

qa⁰tu¹¹

tsɯ²tau³

kɯ²tou³

back

qa⁰ʔdiŋ²⁴

tsɯ²tsʰuɲ³

kɯ²tsʰun³

neck

qa⁰ʑɔ³¹²

tsɯ²zoŋ³

kɯ²zuŋ³

References

Jacques, Guillaume. 2017. On the status of Buyang presyllables: a response to Professor Ho Dah-an. Journal of Chinese Linguistics vol. 45, no.2 (June 2017): 451-457.

Li Jinfang (1999) Buyang Yu Yanjiu. Beijing: Zhongyang Minzu Daxue.

Ouyang Jueya 歐陽覺亞 and Zheng Yiqing 鄭貽青. 1983. Li Yu Jian zhi 黎語簡誌. Beijing: 北京, Chinese Academy of Social Sciences中国社会科学院.

Posted on

刖 ‘cutting the legs below the knee’: a third example of vowel ablaut

After my last two posts discussing words for ‘200’ and for nose-cutting in the context of vowel ablaut, Ma Kun 马坤 sent me a private message inquiring about 刖 *[ŋ]ʷˤ<r>at > ngwaet > yuè ‘amputate the feet’ 0306h, which I had overlooked in my discussion of words of the shape *ŋˤVt having ‘cut off’ semantics. This word turns out to be another example of the spreading process described in my two last post, though only as far as vowel spreading goes. The pharyngealization specification does not spread.

Like 劓 , 刖 refers to the punishment of criminals: it consists of cutting off a criminal’s 跪, OC *[g](r)ojʔ. The verb phrase 刖跪 occurs several times in 《韓非子·內儲說下》and in《說苑. 正諫》, meaning ‘to cut off the 跪’. 跪 is not so much ‘foot’ (Karlgren) as ‘lower leg, leg below the knee’, so 跪 as the verb ‘to kneel’ is derived from the noun ‘leg below the knee’. Proto-Hmong-Mien (Ratliff) borrowed the noun 跪 as *ɟu̯əiH ‘knee’.

Independently from this, we have to modify the vowel of 刖. Bill Baxter observed in 2018 that 刖 seems to rhyme as *-ot in 周易 hexagram 47 ䷮ 困: “九五：劓刖，困于赤紱，乃徐有說，利用祭祀。Accordingly we should now reconstruct 刖 *ŋˤ<r>ot > ngwaet > yuè ‘amputate the leg(s) below the knee’ 0306h.

Thus, like 二  and 皕 (*i), and like 鼻 and 劓 (*i),  刖 and 跪 have the same vowel, *o. Under my hypothesis, there must have existed a nominal compound ‘cutting off the lower leg’, perhaps the name of the punishment, consisting of the bodypart name plus the verb *ŋˤVt ‘to cut off’ whose vowel could be *a or *u, but neither of *i or *o. Thus:

cutting off the nose *m-bi[t]-s + ŋˤVt > ŋit
cutting off the lower legs *[g](r)ojʔ + ŋˤVt > ŋˤ<r>ot

In both cases the first member’s vowel replaced the second’s. However in the last example the pharyngealization feature did not spread. The <r> infix in its ‘distributed action’ meaning  makes perfect sense in a verb for ‘to cut both legs below the knee’.

Posted on

劓 ‘cut the nose’— a second example of OC ablaut

In a recent blog I argued that the rare word 皕 ‘two hundred’ , OC *prik, was based on  百 *pˤrak > paek > bǎi ‘hundred’ 0781a, with the vowel and absence of pharyngealization of 二 *ni[j]-s > nyijH > èr ‘two’ 0564a which must once have preceded 百 within a compound 二 百: that is, that one source of vowel ablaut in MC was the spreading of the vowel and, perhaps, pharyngealization specification of the preceding morpheme within a compound. Here is a parallel example.

There are in the meaning ‘cut off’ three word families in OC whose main syllable began in *ŋ and ended in *t, while the vowel alternates between *a, *u and *i. The reconstruction used below is experimental. It differs little from Baxter and Sagart (2014) in its treatment of rhymes, which is what matters here.

root ŋ(ˤ)at ‘to cut off’

孼 *ŋ(r)at > ngjet > niè ‘concubine’s son’ 0289g
櫱 *ŋ(r)at > ngjet > niè ‘(shoots from) tree stump’ 0289j
枿 *ŋˤat > ngat > niè ‘(shoots from) tree stump’ 0268a
𣦵 *ŋˤat > ngat > ‘fragment of bone (SW)’ 0269a
乂 *ŋat-s > ngjojH > yì ‘mow, cut (v.)’ 0347a
艾 *(mə)-ŋat-s > ngjojH > yì ‘end, to end (K)’ 0347c
𣡌 *ŋˤat > ngat > niè ‘(shoots from) tree stump’ 0252j
歺 *ŋˤat > ngat > è ‘fragment of bone’ 0269b
歹 *ŋˤat > ngat > è ‘bones, flesh removed’ 0269c
刈 *ŋat-s > ngjojH > yì ‘mow, cut’ 0347b
艾 *ŋat-s > ngjojH > yì ‘mow, cut’ 0347c
孽 *ŋ(r)at > ngjet > niè ‘concubine’s son’ 0289g

root ŋˤut ‘cut off; wobbly, unsafe’

㕟 *-k-ŋˤ<r>[u][t]-s > khweajH > kuì ‘cut off (Qie Yun, no text)’ 0534a
髡 *-k-ŋˤut-n > khwon > kūn ‘shave the head’ 0423a
兀 *[ŋ]ˤut > ngwot > wù ‘amputate the feet’ 0487a
扤 *[ŋ]ˤut > ngwot > wù ‘shake, move; shaken, unsafe, endangered (K)’ 0487b
杌 *[ŋ]ˤut > ngwot > wù ‘tree trunk without branches ’ 0487c
阢 *[ŋ]ˤut > ngwot > wù ‘unsafe’ 0487d
卼 *[ŋ]ˤut > ngwot > wù ‘unsafe’ 0487e

root ŋit ‘cut off the nose’

㓷 *ŋ<r>[e]t-s > ngjejH > yì ‘cut off the nose’ 0285e
劓 *[ŋ]<r>i[t]-s > ngijH > yì ‘cut off the nose’ 0537a

The last root is of special interest. 㓷 is a phonetic compound appearing in Shuihudi texts (He Linyi 2:910) and included in the Shuo Wen, but early paleographical and textual attestations are few. While 劓 occurs in the Shang inscriptions and is widely attested textually: so I take 劓 to be the basic form in this root, to which I assign *i. The alternation between *e and *i is unexplained.

The act of cutting off can apply to a large variety of objects: I assume that the alternation between *a, *u and *i here somehow reflects that fact.

Let us suppose that 劓 *[ŋ]<r>i[t]-s  ‘cut off the nose’ continues an older compound whose first member was 鼻 *mV-bi[t]-s > bjijH > bí ‘nose’ 0521c, and the second, either ŋ(ˤ)at or ŋˤut ‘cut off’. Rather than ‘cut the nose’, this would have meant something like ‘nose-cutting’. If, in addition to <r>-infixation and s-suffixation,  we suppose the same process as for 百 *pˤrak and 皕 *prik, here with 鼻 *mV-bi[t]-s ‘nose’ providing the vowel and pharyngealization specification, we obtain 劓 *[ŋ]<r>i[t]-s. This further supports my hypothesis that one source of OC vowel ablaut is the spread within a compound, of the first morpheme’s  vowel and pharyngealization specification, onto the second.

References

Baxter, William H. and Laurent Sagart. 2014. Old Chinese: a new reconstruction. New York: Oxford University Press.

Hé, Línyí 何琳儀. 1998. Zhànguó gǔwén zìdiǎn: Zhànguó wénzì shēngxì 戰國古文字典：戰國文字聲系 [Dictionary of the Warring States script: its system of phonetics]. 2 vols. Běijīng 北京: Zhōnghuá 中華.

 

 

Posted on

皕 ‘two hundred’ and OC vowel ablaut

Some OC lexical doublets have different vowels. Such pairs are sometimes called “ablaut pairs”. What process is behind ablaut is so far not known.

This post discusses a doublet form for the numeral ‘100’, which suggests that one of the processes, at least, behind ablaut, is interaction between the first and second vowels of a lexical compound, such that the first vowel assimilates the second one. Subsequently the compound’s second morpheme individualizes with its new vowel.

Baxter and Sagart (2014) reconstruct the word for ‘hundred’ as 百 OC *pˤrak > MC paek. There is a rare word 皕 ‘two hundred’, not attested before Han. whether in texts or palaeographically as far as I know. The Shuo Wen says it is pronounced like 祕 （pijH < *-it-s), while the Guangyun reads 彼側切 pik. An alternation between MC -it or -ijH and -ik indicates OC *-ik. B&S (2014) did not reconstruct that word but would probably have given OC *prik. The *i vowel in ‘200’ alternates with the *a vowel in ‘100’.  It happens to be the same vowel as that in 二 *ni[j]-s ‘two’. It is possible that *prik arose out of 二百 *ni[j]-s pˤrak, with the vowel of *ni[j]-s  spreading rightward, replacing the original low vowel in the ‘hundred’ morpheme, thus *ni[j]-s pˤrak > *ni[j]-s prik and ultimately 皕 prik, as the vowel sufficed to distinguish 皕 from 百 .

Added Sept 18, 2023: not just the vowel of 二, but the [-pharyngealized] feature appears to have spread from 二 to 皕. This is in agreement with the observation, made in a Chinese PhD some years ago, that in disyllabic words found in early Chinese texts, the two syllables agree in being both type A [+pharyngealized], or type B  [-pharyngealized] . It causes me grief that I have misplaced the author’s name and the dissertation’s reference. I would be grateful if a reader could provide this information. Meanwhile, the pharyngealization harmony in OC disyllabic words, including in the case of 二百 > 皕, can be explained by a process copying the pharyngealization feature of the first word onto the second, overriding the second word’s original specification. This would go some way towards explaining the frequent alternation between pharyngealized and non-pharyngealized cognate words within OC word-families.

Reference

Baxter, William H. and Laurent Sagart. 2014. Old Chinese: a new reconstruction. New York: Oxford University Press. 4

 

Posted on

zhēng yuè 正月 ‘first month’ and the stop/nasal alternation in Old Chinese codas.

One of the sources of the alternation between stop and nasal codas—the ‘陽入對轉’ of  traditional Chinese phonology—is assimilation of a final stop to the nasal initial of the following word. An example is the expression zhēng yuè 正月 ‘first month of the lunar year’. The earliest secure occurrence of 正月 I am aware of, referring to the first month, is in the Luan Shu Fou 䜌書缶, a Zhou bronze inscription:「正月季春元日」’on the first day of the first month’. Palaeographers would probably know earlier ones. The character 正 normally writes *teŋ-s > tsyeng > zhèng ‘correct (adj., v.)’. How did it become used to express the meaning ‘first’  ?

There is in Western Sino-Tibetan (a.k.a. ‘Tibeto-Burman’) a root  for ‘one’ which the STEDT website reconstructs as

#2726 PTB *tyak ⪤ *g-t(y)ik ONE (provisional).

The corresponding Chinese root reconstructs as *tek ‘single, only, only one’. It includes these words: 

隻 *tek > tsyek > zhī ‘single’ 1260c, 

適 *s-tek > syek > shì ‘only’ 0877s, 

啻 *s-tek-s > syeH > shì ‘only’ 0877k.

Root *tek and the root in 正 *teŋ-s differ only in the nasality of the coda. This suggests 正 writes 隻 *tek ‘1’ in a context where its ending had become nasalized. The onset of the following word: 月 *[ŋ]ʷat is likely where nasality comes from. 正月 was an appropriate way of putting *teŋ-ŋʷat < *tek-ŋʷat into readable form.

Reference

Liu Shiru. 1965. Weijin Nanbei chao liangci yanjiu (Studies on Classifiers in Wei-Jin and Nan-Bei dynasties). Beijing: Zhonghua.

Posted on

Chinese 秋 ‘autumn’: time of collecting.

The Chinese name of the autumn: 秋 *tsʰiw > tshjuw > qiū ‘autumn; crop’ is relatable to 揫 *[dz]iw > dzjuw > jiū ‘collect, bring together (K)’. We may have to emend [dz]iw to m-tsʰiw (although Mandarin jiū suggests *m-tsiw). There are other cases of alternation between aspirated and non-aspirated stops/affricates in related words.