It is proposed that oc pharyngealized onset consonants—that is, ‘type-A’ onset consonants—arose out of Proto-Sino-Tibetan plain consonants followed by geminate vowels separated by a pharyngeal fricative. When the first copy of the geminate vowel fell, the initial consonants formed clusters with the pharyngeal fricative, evolving into the oc pharyngealized consonants we reconstruct. In the Kuki-Chin branch of Tibeto-Burman, the pharyngeal fricative fell, and long vowels resulted. This proposal supposes a statistical correlation between Kuki-Chin long vowels and oc type-A words on the one hand, and between Kuki-Chin short vowels and oc type-B words on the other, as originally proposed by S. Starostin. A significant statistic bearing on forty-three probable Chinese-Kuki-Chin cognates supports this correlation. Thus reconstructed, a precursor language of Proto-Sino-Tibetan was aligned with Proto-Austronesian and Proto-Austroasiatic in exhibiting a surface constraint against monomoraic free words: by that constraint, the vowel of an underlying monosyllable was realized as a geminate with an intervening parasitic consonant such as a glottal stop or a pharyngeal fricative, while the vowels of a disyllable remained nongeminate. After reduction of disyllables to monosyllables, this process resulted in a pharyngealized vs. nonpharyngealized consonant distinction in oc.
論文提出對上古漢語咽化聲母（即“A類聲母”）來源的一個假設，認為咽化聲母來源於原始漢藏語的無標記輔音，並且此無標記輔音後跟隨著由咽部擦音[ʕ]分割的雙胞元音（geminate vowels）。在雙胞元音的前半部分脫落後，輔音聲母跟咽部擦音形成了複輔音。新形成的複輔音最終演變為上古漢語的咽化聲母。而在藏緬語庫基-欽（Kuki-Chin）語支中，則是咽化擦音脫落，使雙胞元音變為長元音。本文解釋了漢語與庫基-欽語中一個統計學的顯著相關性：一方面，庫基-欽語含長元音的詞與上古漢語含A類聲母的詞呈顯著相關；另一方面，庫基-欽語含短元音的詞與上古漢語含B類聲母的詞呈顯著相關，正如斯塔羅金（S. Starostin）所說。據此構擬，我們提出原始漢藏語、原始南島語與原始南亞語中的一個共同限制（constraint）：禁止在語流中出現單音拍（monomoraic）詞。在這一限制的作用下，單音節詞的單元音發生了雙胞化（germination），雙胞元音中間插入了喉塞音或咽部擦音之類的次聲輔音。而同時，雙音節詞的元音無變化。漢語經過雙音節詞的單音節化，就形成了無標記輔音與咽化輔音的音位對立。(This article is in English.)
* This is a reworked version of a paper presented at the Recent Advances in Old Chinese Historical Phonology Workshop held in London on 5 and 6 November, 2015 as part of the European Research Council Grant ‘Beyond Boundaries: Religion, Region, Language and the State’. We thank the participants at the workshop for their comments, and the participants in an Academia session on the same paper for their useful discussion, particularly South Coblin, Zev Handel, Doug Henning, Guillaume Jacques, Johann-Mattis List, Thomas Pellard, Adam Smith, Miguel Carrrasquer Vidal, Gong Xun, Lukáš Zádrapa. The first author’s work is part of the program Investissements d’Avenir, overseen by the French National Research Agency, anr-10-LABX-0083 (Labex Empirical Foundations of Language).
1 The Reconstruction of oc Pharyngealized Consonants
Old Chinese main syllables fall into two groups, called by Pulleyblank (1977–78) “type A” and “type B”. With minor adjustments, type-B syllables are those traditionally classified as “division-iii” (sānděng 三等). In Karlgren’s reconstruction (e.g., in Karlgren 1954), type-B syllables are reconstructed with a yod *-i̯- before the main vowel, and type-A words lack this yod; Li Fang-kuei (1971) and Baxter (1992) adopted the same solution, using the notation *-j- instead of *-i̯-. But this reconstruction is now widely regarded as unsatisfactory, and several alternative analyses have been proposed (see Baxter & Sagart 2014:68–80 for details).
Inspired by the treatment in Norman (1994), Baxter & Sagart (2014) assign pharyngealization to oc type-A words, and absence of pharyngealization to type-B words. The main argument for reconstructing pharyngealization is a set of sound changes that occurred during the Hàn period (206 bce–220 ce), which affected type-A syllables and type-B syllables differently: original high vowels remain high in type-B syllables, but are lowered in type-A syllables; and original low vowels, which are raised in certain environments in type-B syllables, remain low in type-A syllables. Also, initial consonants often underwent palatalization in type-B syllables, but escaped such palatalization in type-A syllables. Reconstructing pharyngealization in the onset of type-A syllables seems to provide a plausible explanation for these differences, more so than any of the alternative proposals.1 Accordingly, Baxter & Sagart (2014) reconstruct Old Chinese with two parallel series of initial consonants, nonpharyngealized and pharyngealized (Table 1).2
The sound changes supporting the reconstruction of pharyngealization are rather late as the Hàn period is generally considered to be at the end, or even after the end, of the Old Chinese period. Having no earlier direct evidence for the phonetic nature of the A/B distinction, we project pharyngealization all the way back to the earliest Old Chinese.
However, as Baxter & Sagart observe (2014:73–74), such a phonological system is typologically unusual: languages with pharyngealized consonants generally have more nonpharyngealized consonants than pharyngealized ones. We think of typologically unusual features as short-lived and unstable, though not necessarily impossible. It is entirely possible that the onset consonant inventory in Table 1 existed only for a short period of time at the end of the oc period.
2 A Model of the Origin of oc Pharyngealization
Outlined below is a hypothetical model of how such a contrast may have arisen. Briefly, we propose that the pharyngealized consonants arose out of clusters of plain consonants followed by a pharyngeal fricative segment [ʕ], that is: *CʕV- > CˤV-. We assume that such *Cʕ- clusters would have existed for all Old Chinese onset consonants.3 At a still earlier stage, we derive these *CʕV- onsets from strings in which two copies of a vowel flanked [ʕ]: *Caʕa-, *Ciʕi-, *Cuʕu- etc., or in other words, a geminate vowel interrupted by a pharyngeal fricative. The *Cʕ- clusters, and ultimately the pharyngealized onset consonants reconstructed, were formed when the vowel’s first copy fell, as in Table 2.
3 Lushai (Mizo) and Kuki-Chin Vowel Length and oc Pharyngealization
Starostin (1989:327sq) made an observation that may be relevant to the origin of the Chinese A/B distinction. He argued that the Chinese A/B distinction was correlated with vowel length in Lushai (also known as Mizo), a Tibeto-Burman language of the Kuki-Chin group, as follows:
- 1.Lushai long vowel : oc type A
- 2.Lushai short vowel : oc type B
Assuming that this correlation is valid, how could one account for it?4 If the Proto-Sino-Tibetan precursor of correlation 1 above was *CViʕVi-, it is easy to derive the Lushai vowel length distinction simply through loss of the pharyngeal segment (Figure 1).
As for the origin of sequences like pst *CViʕVi-, we propose that they could arise from a constraint barring monomoraic free forms, for which there are typological parallels in both Austronesian and Austroasiatic.
4 Bimoraicity in Austronesian
The reconstructible free content morphemes of Proto-Austronesian overwhelmingly consist of words of two syllables or more. Meaning-associated monosyllables exist too, but either reduplicated, or as formatives at the end of longer words: Austronesianists call them ‘roots’ (for collections of roots see Blust and Trussell, ongoing; Wolff 2010). For instance, root *-sek ‘cram, crowd’ (Blust) occurs in pan *seksek ‘cram in’, pmp *hasek ‘jam, cram, crowd’, *be(n)sek ‘overcrowded’, pwmp *dasek, *desek ‘compress’; as well as in words attested in a single language, e.g. Banjarese barasak ‘overcrowded’: but simple monomoraic monosyllables straightforwardly derived out of root *-sek ‘cram, crowd’ cannot be found. The fact that the root is preserved in recognizable form at the end of words in *hasek, *be(n)sek etc., while nonfinal elements in compounds are phonologically reduced, suggests that pan was stress-final. In the examples above, the highly eroded lexical material preceding the root: *ha-, *be(n)-, *da-, *de-, bara- is typically not recognizable. Evidently roots like *-sek once existed as free morphemes, without reduplication or compounding, and have been largely driven out of the modern languages. Two languages: Bunun (Taiwan) and Cebuano (Philippines) preserve some of them: their phonetic shape in these languages indicates that as free morphemes they were realized with a geminate vowel interrupted by a glottal stop, presumably to satisfy a constraint against monomoraic free forms. For instance, Wolff (2007) argues that Cebuano suʔúk ‘remote corner, being deep inside’ is the regular reflex of the root *-sek. Bunun retains more of these ‘stretched’ monosyllabic roots as free forms, with geminate vowel and optional glottal stop (Table 3).
Personal pronouns also show traces of this alternation. The root *-Su ‘2sg’ occurs unchanged as the last syllable of di- or polysyllabic forms Thao ihu ‘thou’, Atayal isuʔ ‘you’, Amis k-iso ‘you’, Kanakanabu ii-kasu ‘thou/you’; as a free form, however, it exhibits vowel gemination in Bunun su(ʔ)u ‘you’, Saisiyat shoʔo ‘you’, Tsou súu ‘thou’. Similarly, the root *-mu ‘2pl’ occurs unchanged in Atayal simu ‘you and he/she’, Saaroa iɬa-mu’you’, Bunun a-mu ‘you (pl.)’, Amis ka-mo ‘you (pl.)’, Kavalan i-mu ‘you (pl.)’, but has vowel gemination in the free forms Bunun mu(ʔ)u ‘you (pl.)’, Tsou múu ‘you (pl.).
5 Bimoraicity in Austroasiatic
A similar constraint against monomoraic free nominals was identified in Proto-Austroasiatic by Zide (2002), who termed it ‘the bimoraic constraint’. The strategies used by Austroasiatic and Austronesian speakers to make monosyllables compatible with the bimoraic constraint are very similar. This can be illustrated by Munda forms from Anderson (2004) in Table 4.
In Table 4 we see monosyllabic roots for ‘hand’ and ‘foot’ made bimoraic by reduplication in Gutob, by the prothesis of i- in Juang, and by the gemination of the vowel with insertion of a glottal stop in Gorum. The Sora forms appear to have lost a vowel before the glottal stop: this makes them similar to Old Chinese type A forms, stage 2, in Table 2.
6 oc Type A/B as Bimoraicity
We propose that the A/B distinction, in the form that we attribute to Proto-Sino-Tibetan (Figure 1), continues an even earlier contrast in a language ancestral to Proto-Sino-Tibetan, where the same bimoraic constraint observed in Proto-Austroasiatic and in Proto-Austronesian existed. In that language, type-B forms would have had two syllables, and type-A forms would have had the ‘stretched monosyllable’ *CViʔViC structure found in Bunun and in Gorum.5 For instance, the oc pair type-B 入 *n[u]p ‘enter’ and type-A 內 *nˤ[u]p ‘bring or send in’, which at some stage contained the same monosyllabic root,6 would go back to pst type-B #nup vs. type-A #nuʕup; these in turn would go back to pre-pst disyllabic #σ-nup (‘σ’ = ‘syllable’) vs. type-A #nuʔup, a stretched monosyllable. The bimoraic constraint can be formulated so as to prohibit not just free forms of one mora, but free forms with an odd number of morae (assuming that the bimoraic constraint applies to feet, and that words must have an integral number of feet). Under this formulation, if the language satisfied the constraint by ‘stretching’ both monosyllables and the last syllable of trisyllabic free forms, then pst *nuʕup could go back to either of pre-pst *nuʔup or *σ-σ-nuʔup. Evolution into Proto-Sino-Tibetan would then involve (1) reduction of non-final syllables, and (2) change of intervocalic glottal stop to pharyngeal fricative *-VʔV- > *-VʕV-, in circumstances that have not been established.
7 Testing the Correlation between Kuki-Chin Vowel Length and oc Type A/B
We compare the Old Chinese A/B distinction in the reconstruction of Baxter-Sagart (2014) with vowel length, not in Lushai, but in Kuki-Chin, the subfamily of Tibeto-Burman to which Lushai belongs, using VanBik’s reconstruction of Proto-Kuki-Chin (VanBik, 2009). The advantages of using reconstructed Proto-Kuki-Chin are twofold: (1) the reconstructed forms are a clearly circumscribed set, selected in advance for another purpose, and choosing examples from the extensive Lushai lexicon might introduce a selection bias; and (2) in the reconstructed forms, any single-language irregularities are likely to be ironed out.
Null hypothesis: there is no positive correlation between pkc long vowel and oc type A on the one hand, and between pkc short vowel and oc type B on the other hand.
The Proto-Kuki-Chin material was scanned to identify a list of likely Chinese cognates, which are listed in the Appendix. Although the precise correspondences between Old Chinese and Kuki-Chin are not fully understood, we were guided by known regularities of Old Chinese – Tibeto-Burman comparison and by Kuki-Chin phonological history: for example, pkc *th- appears to correspond regularly to Old Chinese *s-. Certain kinds of Kuki-Chin forms were omitted because the vowel length cannot be determined unambiguously. Because Kuki-Chin verbal morphology affects vowel length in ways that are only partly understood, the purview was limited to nominals: VanBik provides a list of those (2009:451–515). We also excluded open-voweled syllables, as there is no length contrast: all syllables with open vowels are long in Proto-Kuki-Chin (VanBik 2009:323). In addition, there is no length contrast among vowels preceded by /i/ or /u/, therefore these were excluded as well. The remaining forms are all cvc or cvvc monosyllables. Also excluded from comparison are:
- • pkc words with long and short variants, e.g. ‘elbow’ *ki(i)w 3, ‘egg’ *ɗu(u)y 4, *tu(u)y 4, ‘yard, armspan, cord’ *la(a)m 4;
- • oc words with A/B variants, e.g. 入 *n[u]p ‘enter’ and 內 *nˤ[u]p ‘bring or send in’; 糲 *[r]ˤat and *[r]at-s ‘dehusked but not polished grain’
- • oc words of uncertain type, such as 髟 *s(ˤ)ram ‘long hair’;7
- •probable loanwords: ‘silver’, pkc *ŋuun, oc 銀 *ŋrə[n]8
- •comparisons requiring large semantic shifts: ‘pig’, pkc *wok 3 vs. 富 *pək-s > pjuwH > fù ‘rich; wealth’.
Forty-three comparisons in total were retained (see Appendix). A 2 x 2 contingency table was built from the words in the four possible categories (long-type A; long-type B; short-type A; short-type B), as follows in Table 5.
The P value, or significance of the deviation from the null hypothesis, was calculated from Table 5 using Fisher’s exact test.9 Application of the test returns the P value 0.032. A P value under 0.05 is usually regarded as significant in scientific works. The P value 0.032 found here means that the probability under the null hypothesis of finding the same—or a smaller—proportion of forms that violate the proposed correspondence (long vowel ↔ tone A, short vowel ↔ type B) as in Table 5 is approximately 3.2%. In other words, on present evidence the null hypothesis should probably be rejected: there probably exists a positive correlation between Kuki-Chin long-short and Old Chinese A/B, as Starostin proposed. We are aware that this result could be different if other cognates were identified. It may be necessary in future to modify our list and to calculate the P value again. We invite replication or falsification of our results by other investigators on the basis of independently established cognate lists. Another avenue for replication or falsification would be to examine cognates of Old Chinese type A and B words in different Kuki-Chin languages like Tiddim or Lushai, for which significant lexical lists exist; or in Proto-Northern Chin (Button 2011).
8 Explanation of Mismatches
Eleven comparisons in Table 5—almost one in four—behave contrary to prediction; they are italicized in the Appendix. The explanation for these mismatches is probably due to dealing not with exact cognates but with free forms having odd versus even numbers of syllables, yet both ending in the same monosyllabic root. For instance, pkc *raʔ ‘fruit’ and oc 糲 *[r]at-s ‘dehusked but not polished grain’ would both go back to the monomoraic final syllable of a pre-pst disyllable #σ-ras, while oc *[r]ˤat would go back to the pre-st stretched monosyllable #raʔas. Thus this explanation could account both for Chinese forms with A/B variation and for long-short variants in Kuki-Chin.
Anderson Gregory D. S. Norman Zide . 2002. Issues in Proto-Munda and Proto-Austroasiatic nominal derivation: The Bimoraic constraint. In Papers from the 10th Annual Meeting of the Southeast Asian Linguistics Society, ed. Macken Marlys , 55–74. Tempe: South East Asian Studies Monograph Series, Arizona State University.
- Search Google Scholar
- Export Citation
. Anderson Gregory D. S. Norman Zide 2002. Issues in Proto-Munda and Proto-Austroasiatic nominal derivation: The Bimoraic constraint. In Papers from the 10th Annual Meeting of the Southeast Asian Linguistics Society, ed. , Macken Marlys 55– 74. Tempe: South East Asian Studies Monograph Series, Arizona State University.
Blust Robert and Trussell Steven . The Austronesian Comparative Dictionary (accessed 26 October 2016. http://www.trussel2.com/acd/).
Karlgren Bernhard . 1954. Compendium of phonetics in Ancient and Archaic Chinese. Bulletin of the Museum of Far Eastern Antiquities 26:211–367.
Starostin Sergei A. 1989. Rekonstrukcija drevnekitajskoj fonologičeskoj sistemy [A reconstruction of the Old Chinese phonological system]. Moscow: Nauka, Glavnaya Redakcija Vostočnoj Literatury.
VanBik Kenneth . 2009. A Reconstructed Ancestor of the Kuki-Chin Languages. STEDT Monograph Series Number 8. Berkeley: University of California.