Some Attempts at Enhancing Old Chinese Reconstructions Through the Lens of Palaeography

1 Palaeographic Material as Sources for Old Chinese Phonology

In his 2016 review of the latest reconstruction system of Old Chinese (OC) by Baxter and Sagart (2014) (hereafter, BS),¹ Christoph Harbsmeier (2016, 498) concludes that “we need reconstructions, no matter how unreliable these may prove to be”. Harbsmeier’s words were probably too harsh, since, after almost a decade, the system proposed by BS still remains ‘the state of the art’, and many of its features have proven to be essentially correct. The case of vowel purity is exemplary: a comparison between BS and other reconstruction systems regarding the medial vowels in rhyming words of the Shijing 詩經 (Book of Odes, eleventh-seventh century BCE)² demonstrates significantly superior performance of the former, indicating a much more well-ordered pattern of rhyme schemes than any previously proposed system (List et al. 2017, 1‑17).

The methodology underlying BS’s reconstruction has been effectively synthesised by Hill (2019, 188‑94), and I will largely draw on this schematic and punctual synthesis for many of the instances discussed in this paper. In a nutshell, the methodology goes as follows: it starts from the Middle Chinese reading of a character as given in the rhyme dictionaries and rhyme tables,³ and projects it back to OC by taking into account both language-internal and language-external data. To the first category belong both philological sources such as the Shijing rhyme patterns, ancient glosses, and the xiesheng 諧聲 series,⁴ but also systematic accounts of Sinitic varieties not descendants of MC (e.g. Min 閩) and word families alternations. The second category mainly includes older layers of loans from OC into neighbouring languages (e.g. Lakkia and Rục, among others).⁵ Palaeographic data also play a role in the BS system, although they are not employed systematically.

Through some illustrative examples, I will argue that a further step should be included as an integral part of the whole methodology: the systematic verification of reconstructed items against the numerous cases of phonetic borrowings detectable in pre-Han excavated manuscripts. This claim is far from being unprecedented. Many scholars have long recognised the potential that palaeographic data could hold for phonological reconstructions (e.g. Qiu 1979; Liu 2005; Ye 2017), and the same could be said for the reverse (e.g. see Jacques 2022, 488‑91). The significant number of texts unearthed through archaeological discoveries, and unfortunately, also through the occurrence of looting ancient tombs in recent decades, now provides a vast amount of data deserving a major role in research on Old Chinese phonology.⁶

To this day, the most remarkable attempts to refine the BS system through the contribution of palaeography have been made by Haeree Park (2016, 31‑72). Some attempts at testing rhyme class distinctions in different phonological reconstructions – also including the BS system – through palaeographic evidence also exist in Chinese-language scholarship (e.g. Yang 2022). Nonetheless, Chinese historical phonology and Chinese palaeography remain by and large two fairly distinct fields with only sporadic interaction between them.

The examples discussed in this brief paper will focus on phonetic borrowings,⁷ attempting to show how they can prove to be essential evidence for at least three different questions, namely a.) reconstructing items currently not included in BS, b.) disentangling merged rhymes, and c.) questioning and verifying reconstructed forms in BS.

2 Reconstructing Items not Included in BS

One of the most evident cases in which palaeographic material can help linguists in enhancing current reconstructions of OC are those characters that lack a reconstructed form but function as clear instances of phonetic loans in the manuscripts. For example, consider the case of the sentence final particle yi 殹. This particle is rather rare in transmitted texts while it is frequently seen in unearthed sources. Numerous studies, especially in Mainland China, have long argued that yi 殹 should be understood as a sort of Qin 秦 region counterpart of the more common sentence final nominal predication marker ye 也 (e.g. Zhang 2011, 570‑85). Yet, the retrieval of new caches of bamboo slips from the Chu 楚 region has revealed that the particle was frequently used by Chu scribes as well.⁸

Reconstructing this item can prove to be challenging in some ways: the character never appears in the Shijing and all the other items included in the same xiesheng series have no reconstructed form in BS. This means that its MC form retrievable from the Guangyun cannot be projected back to OC.⁹ The MC form ‘ejH (Guangyun 1958, 375)¹⁰ could possibly go back to different rhymes and different initials in OC: the rhyme -ej 齊 can originate from OC *-ij-s (zhi bu 脂部), *-it-s (zhi bu 質部) or *-et-s (ji bu 祭部) and the initial consonant ‘- 影 can go back to OC *q- or *ʔ-.

A significant piece of data for assessing the correct reconstruction comes from the collection of bamboo manuscripts at Anhui University (hereafter, ADJ1 or ADJ2 in relation to the edited volume): on slip no. 40 of the Anhui University Shijing manuscript,¹¹ there is a line which has a perfect match in the transmitted text of Shijing 24/11/18.¹² In this line the character yi 殹 clearly functions as a loan for the particle yi 伊 *ʔij (zhi bu 脂部):

ADJ1: 隹絲殹緍

Shijing 24/11/18: 維絲伊緡

It is silk that makes the thread.¹³

13  Transl. by Meyer, Schwartz 2022, 199.

Another instance of phonetic borrowing is found on slip no. 9 of the Shanghai Museum manuscript (SBJ2) Zigao 子羔,¹⁴ where yi 殹 stands for the particle yi 抑 *[ʔ](r)ik.

SBJ2: 殹(抑)亦成(誠)天子也與(歟)?

Yet perhaps they truly were the sons of Heaven?!¹⁵

15  Unless otherwise stated, all translations are by the Author.

The example in the Anhui University Shijing allows us to opt for the rhyme *-ij-s (zhi bu 脂部) in OC. Yet, to account for the example in the Shanghai Museum manuscript, we could hypothesise a final glottal stop (*-ʔ) in the reconstruction. As for the initial consonant, no contact with k- 見, g- 群, h- 匣, x- 曉 or y- 以 can be detected in the xiesheng series, suggesting that the MC initial consonant ‘ 影 should be projected back to the OC initial *ʔ. The character entry yi 殹 can thus be reconstructed as OC *ʔˤi[j](ʔ)-s.

A similar case occurs with the character yan 牪. This character is rather rare: it lacks an entry in the Shuowen jiezi 說文解字 (Explain the Graphs to Unravel the Written Words, second century CE),¹⁶ never occurs in the Shijing, has no xiesheng series, and no entry in the Guangyun. This makes its reconstruction quite hard. Its MC reading, though, can be retrieved from the Yupian 玉篇 (Jade Chapters, sixth century CE in Kangxi zidian 康熙字典 2015, 651) where the fanqie 反切 niu juan 牛眷 MC ngjuw kjwenH is given, yielding the MC ngjwenH.

The rhyme -jwen 仙 can be reconstructed as OC *-ron or *-ror (yuan bu 元部), while the initial consonant ng- 疑 can be reconstructed as *m-q^h-, *m-G-, *N-q^h-, *N-G- when showing connections with MC k- 見, y- 以 or ‘- 影, or as OC *ŋ-.

Also in this case, two instances of phonetic borrowings in the manuscripts can help us test what has been proposed so far: the character yan 牪 appears in two pericopes of the two manuscript versions of Cao Mo zhi zhen 曹沫之陳 (Cao Mo’s Battle Formations) from the collections of the Shanghai Museum and Anhui University. The editors of the Anhui University manuscript have interpreted the character as a phonetic borrowing for the words yuan 願 *[ŋ]o[n]-s (to desire) and juan 倦 *[g]ro[n]-s (fatigued) (ADJ2 2022, 53, 66), while the editors of the Shanghai Museum manuscript (hereafter, SBJ4) only agree on the second reading (SBJ4 2004, 267‑8):¹⁷

ADJ2 slip no. 3: 牪(願)尔正 (功)

SBJ4 slip no. 37: 牪(願)尔正(功)

May you want to rightly apportion merits.

ADJ2 slip no. 24: 不可田(使)牪(倦) = (倦)則不行

SBJ4 slip no. 38: 不可思(使)牪(倦) = (倦)則不行

Do not let [those in command of the army] get fatigued or they will no longer be able to march.

These readings, acceptable with respect to the broader contexts of the pericopes, are both supported by phonological evidence and, at the same time, allow us to enhance our phonological understanding: if yan 牪 here effectively functions as a loan for yuan 願 *[ŋ]o[n]-s or juan 倦 *[g]ron-s, it would mean that it can be rather safely reconstructed as *[ŋ]ro[n]-s.¹⁸

In the absence of a xiesheng series for further investigation, a reconstruction with a nasal prefix – be it *m- or *N – and an uvular as onset could also be possible. As the Yupian glosses the character as related to niu 牛 (ox),¹⁹ we could speculatively opt for a *m- prefix, frequently attested in animal names. However, the reconstruction with *[ŋ] onset remains the most plausible in light of the phonetic borrowing examples detected in the manuscripts.

3 Disentangling Rhymes that Have Merged

In the centuries separating the language of Mengzi 孟子 (fourth-third century BCE) from the language of the Qieyun 切韻 (601 CE) many rhymes coalesced. Our current understanding of the phonological developments of Chinese allows us to disentangle many of these cases when projecting MC readings back to OC. However, in some cases such disentangling is subject to more reservations. In this respect, the phonological data encoded in the early script attested on the unearthed texts can once again prove to be extremely valuable.

This is the case of the rhymes *-at and *-et (yue bu 月部) which have the same reflexes in MC division III after acute initials, making them difficult to disentangle when reconstructing OC (Schuessler 2009, 225). I take the character shi 筮 MC dzyejH (achillea stalk, divination by achillea), reconstructed as *[d][e][t]-s by BS, as an example.

This character appears twice in the Shijing but it does not rhyme (Baxter 1992).²⁰ Fortunately, it is attested on many unearthed manuscripts from the Chu region where it functions as a phonetic borrowing. One of its best-known occurrences is in the Mawangdui 馬王堆 silk manuscripts (hereafter, MWD4) of the Laozi 老子 (late third century – early second century BCE) where it functions as a phonetic loan for the word shi 逝 *[d]at-s (to leave, to go afar) in a line which has a parallel in the stanza no. 25 of the received edition (MWD4 2014, 42, 206).²¹ The retrieval of different excavated versions of the Laozi allows us to make a deeper survey of the ways this word was written in early manuscripts: the bamboo manuscript preserved at Peking University (second-first century BCE) (hereafter, BD2) writes the character di 慸 (lofty, extreme),²² which is likely to function as a phonetic loan as well (BD2 2012, 156).²³ Its phonophore dai 帶 is reconstructed as *C.tˤa[t]-s, making it a good candidate for phonetic borrowing. Highly informative is also the graph attested on slip no. 22 of the Laozi bamboo manuscript A (late fourth century BCE) found at Guodian 郭店 (hereafter, GD):²⁴ , usually transcribed as .

At the time of its first attestations, this complex graph sparked a heated debate among palaeographers and philologists both on its transcription and on its reading. Some scholars suggested the character stood for a lexical variant in the line of the Laozi, while others suggested it merely represented a phonetic borrowing for the reading shi 逝 of the received version.²⁵

The debate on how to appropriately read it seems to be settled: the character frequently appears in later discovered caches of bamboo manuscripts where it usually stands for the reading shi 逝, confirming that the Guodian version of the Laozi did not show any lexical variation or innovation with respect to the received text. As for its transcription, I hypothesise that its phonophore should be identified in the component , which is the most stable part of all attested graphic variants.²⁶ He Linyi (1999, 196‑204) and Liao Mingchun (2003, 222) identify the phonophore as xue 辥 *s.ŋat. The substitution with the phonophore zhe 折 *N-tet is only visible in Han dynasty documents (see Li 2012, 115).

Receptus: 大曰逝 *[d]at-s

GD: 大曰 (辥 *s.ŋat ?)

MWD4: 大曰筮 *[d][e][t]-s

BD2: 大曰慸 *C.tˤa[t]-s

What is great goes afar.²⁷

27  Transl. by Ryden 2008, 53.

The evidence coming from the Laozi manuscripts seems to suggest that reconstruction in *-at for the character shi 筮 could possibly be more felicitous.²⁸

4 Verifying Reconstructed Items

Another common case in which palaeography turns out to be a valuable tool for questioning or verifying phonological reconstructions is when we come across some ‘mismatches’ between the reconstructed form and the phonetic information encoded in the ancient graphs, or when the ancient graphs highlight some phonological – and sometimes even etymological – connections no longer visible in the kaishu 楷書 script.

The instance of feng 風 (wind) and fan 凡 (all) is exemplary. Though uneasy to tell by simply looking at their kaishu forms, both graphs actually share the same phonophore, as also confirmed by Xu Shen 許慎’s gloss of the entry feng 風 in the Shuowen jiezi.²⁹ This is extremely evident when it comes to the Chu script, where feng 風 is generally written as , while fan 凡 is usually written as . However, Baxter and Sagart reconstruct the two items with different rhymes: feng 風 *prəm and fan 凡 *[b]rom. This ‘mismatch’ seems problematic, since it contradicts the basic assumption of Chinese Phonology – formulated already by Duan Yucai 段玉裁 (1735‑1815 CE) – that “characters with the same phonophores must have belonged to the same rhyme class in ancient times” (tong sheng bi tong bu 同聲必同部).

Baxter transcribes the MC readings of the two characters as pjuwng and bjom. The transcription of the latter, however, could eventually be questioned: fan 凡 is given the fanqie fu fan 符䒦 MC bju phjomH in the Guangyun (1958, 233), but the character fan 䒦 – indicating its rhyme – is actually spelled with the fanqie fu fan 孚梵 MC phju bjomH or phju bjuwng (Guangyun 1958, 548). Interestingly, the second reading of the character fan 梵 matches the rhyme of feng 風 (-juwng 東) and would thus allow to reconstruct the same rhyme also in OC.

The -juwng 東 rhyme after labial initials either goes back to an OC rounded vowel or is reconstructed as OC *-əŋ (zheng bu 蒸部). Consequently, the two words could possibly be reconstructed as *p[o, u]ŋ and *b[o, u]ŋ or as *pəŋ and *bəŋ, which seem to better match the way they were once written on bamboo texts. However, this is not supported by the rhyming patterns of the Shijing, where the character feng 風 is clearly found as rhyming in *-əm. It is thus likely, that in this case the same phonophore was used to represent two non-perfectly identical syllable types, in the lack of a better choice (see Baxter, Sagart 2014, 66).³⁰ Furthermore, though it is true that the fanqie spellers can sometimes seem ambiguous as having themselves multiple readings, it is also true that each fanqie speller is generally only employed with its more frequent and basic reading (Pulini, List 2024).

Cases of ‘mismatches’ between reconstruction and phonetic borrowings in early texts can also prove helpful for reconstructing OC pre-syllables: it is the case of shan 善 *[g]e[n]ʔ (to be good), which, though being a core term in many early Chinese texts, never rhymes in the Shijing. Baxter and Sagart (2014, 78) have identified some evidence for reconstructing a velar initial. However, this reconstruction can be questioned on the grounds of a phonetic loan in the Cao Mo zhi zhen manuscripts (ADJ2 slip no. 11; SBJ4 slip no. 18) where the character chan 纏 *[d]ra[n] clearly functions as a loan for the verb shan 繕 (to repair), which takes shan 善 as its phonophore.

e.g. ADJ slip no. 11: 纏(繕) (甲)利兵

e.g. SBJ slip no. 18: 纏(繕) (甲)利兵

Maintain armours and weapons in good conditions.³¹

This example could possibly be explained by reconstructing shan 善 with a *k- pre-initial before a *d- onset.³² This would account both for the evidence found by Baxter and Sagart for reconstructing a velar in the onset and for the suitability of the character chan 纏 to function as a loanword for shan 繕 in excavated texts. Some further data, also coming from neighbouring languages, could prove useful for further validating this hypothesis.

The last instance of ‘mismatch’ to be explored in this paper is linked to a person’s name: Cao Mo 曹沫, a general of the State of Lu 魯 active in the seventh century BCE, after whom the two aforementioned Cao Mo zhi zhen manuscripts from the collections of the Shanghai Museum and Anhui University are named.

The case of Cao Mo is particularly interesting since both the manuscripts and the transmitted sources show a remarkable degree of graphic variation in relation to this name. In the transmitted literature, the general of Lu is called Cao Mo (either written as 曹沫 or 曹昩), Cao Gui 曹劌 or Cao Hui 曹翽.³³ These different graphic variants all seem to point to the meaning of something ‘obscure’, ‘indistinct’, ‘blind’. This meaning is shared by the characters mo 昩 and gui 劌,³⁴ of which the characters mo 沫 and hui 翽 are likely to be allographs. Furthermore, apart from being semantically related, the two variants of the name of the Lu general attested in the transmitted literature also share the same rhyme (dieyun 疊韻) *-at (yue bu 月部).

Conversely, the manuscripts show the graphic variants and for the first character and the graphs ,, , and for the second one.

By looking at the different graphic variants we can isolate two different phonophores for the first character: cao 曹 *N-tsˤu and gao 告 *kˤuk-s; and four phonophores for the second one: mo 末 *mˤat, sui 歲 *s-qʷʰat-s, mie 蔑 *mˤet and wan 萬 *C.ma[n]-s. A glimpse through the reconstructions in BS clearly reveals a rather complex picture alternating several different rhymes for one single person’s name. This requires further investigation.

It is likely that the phonophore in the graphs and attested on the Cao Mo zhi zhen bamboo slips is not solely the component gao 告, but rather . The latter could possibly be identified as an allograph of zao 造 *[dzˤ]uʔ, which matches the reading cao 曹 *N-tsˤu quite well. Among the early graphic variants of zao 造, in fact, we find the graph (jicheng 集成 11023), whose semantic component 戈 is sometimes found to be equivalent to the signific 攵 of the graph (synonymous significs, Park 2016, 313). This hypothesis is further supported by the Qinghua *Chi zheng zhi dao 治政之道 (The Way of Rectifying the Government) manuscript slips nos. 34, 37 (see QH9), where the graph is in fact used as zao 造.

The character zao 造 *[dzˤ]uʔ is said to take itself gao 告 as its phonophore in the Shuowen jiezi.³⁵ Nonetheless, Chen Jian (2006, 100) has clearly shown how the component gao 告, already found in bronze inscriptions, is the graphic evolution of an early component cao 艸 *[tsʰ]ˤuʔ functioning as phonophore.³⁶

As for the second character Mo, the picture is much more complicated: the spelling with the phonophore sui 歲 *s-qʷʰat-s could be subsequent to the evolution of the cluster *k.m into kw, a phenomenon attested in the evolution of other words from OC to the MC stage: e.g. 袂 *k.mˤet > MC kwet (Sagart, Baxter, Beaudouin 2022). This usage suggests a late datation of this spelling and accounts for its absence in the Cao Mo zhi zhen manuscripts.

The manuscript version of the Sunzi bingfa 孫子兵法 (Master Sun’s Art of War) excavated at Yinqueshan 銀雀山 (hereafter, YQS) helps us in dating this evolution to have taken place at least by the early second century BCE (terminus ante quem).³⁷ General Cao Mo is in fact mentioned in Sunzi A11/12/5 and his name is attested on YQS slips written with the phonophore sui 歲 *s-qʷʰat-s (YQS 1985, 21).

As for the character mie 蔑 (tiny, discard, reject, poor sighted), it cannot be associated with an *-at rhyme if we base our investigation uniquely on its Guangyun spelling (i.e. met) (Guangyun 1958, 496). Nonetheless, it is interesting to notice that some characters within the same string of homophones of mo 末 MC mat given in the Guangyun actually share mie 蔑 as phonophore: mie , mo , mo 瀎, mie (Guangyun 1958, 497). This implies that indeed these entries can be reconstructed with an OC *-at rhyme. Furthermore, a gloss preserved in the Xiao Erya 小爾雅 (Little Erya late third century BCE) and a gloss by Duan Yucai to the Shuowen jiezi might provide a clue. The entry mie 蔑 is said to be “the same as mo 末” (mo ye 末也) in the Xiao Erya (in Kangxi zidian 2015, 1410), while Duan Yucai states that this is “an allograph of mo 眜” (tong zuo mo 通作眜 Shuowen jiezi 1983, 145), another character meaning ‘to be poor-eyed’.

These glosses, along with the usage attested in the manuscripts, suggest that the character mie 蔑 once also had a reading *mˤat, which had already been lost by the time of composition of the Qieyun, but was retained in some characters which take it as their phonophore.

As for the character wan 萬 (ten-thousands), – which also has a mo reading in Mandarin, generally employed in surnames (e.g. Moqi 万俟) – it matches the OC *-at rhyme of the other three phonophores only in the medial vowel, but not in the final consonant, though both *-n and *-t are acute codas. The phonophore wan/mo 萬 is thus in a dui zhuan 對轉 relation with the other three phonophores employed for the name of the Lu general.³⁸ Yet there are some characters within BS reconstruction that though having wan/mo 萬 as a phonophore, are reconstructed as *-at: e.g. mai 邁 *mˤrat-s.

Interestingly, the very well-known allograph wan 万 is usually written in the manuscripts as mian 丏 , which is glossed as bu jian 不見 (to be blind, unable to see) in the Shuowen jiezi (1983, 423), resembling closely the meaning given for the entries mie 蔑 and mo 眜, both indicating a poor sight.³⁹

e.g. GD Tang Yu zhi dao 唐虞之道 (The Way of Tang and Yu) slip no. 27: 丏 (万)勿 (物)(皆)訇 (暗)

The Ten-Thousand beings are all obscure.

Regardless of the scenario, the use of mie 蔑 and wan/mo 萬 as phonophores for a *mat syllable type continues to serve as compelling evidence that requires careful examination. This intriguing instance of phonetic borrowing undoubtedly merits further investigation and discussion.

5 Conclusive Remarks

By looking at some informative examples, this paper has tried to explore the potential of systematically including palaeographic sources as an integral part of the methodology underlying current reconstructions of OC focusing on the phenomenon of phonetic borrowing, which is extremely frequent in early texts.

As the examples discussed have shown, the phonetic loans attested in early manuscripts can sometimes prove to be an essential tool both for reconstructing items ex novo, as well as for questioning and verifying some OC forms currently included in phonological reconstructions. On the other hand, a deep understanding of OC phonology turns out to be an irreplaceable tool for palaeographers and philologists to correctly transcribe and read early sources. A more porous boundary between these fields is clearly needed. This paper has tried to take a modest step in this direction.

Bibliography

Primary Sources (excavated)

Primary Sources (transmitted)

Secondary Literature