Reconstructions of Old Chinese

Reconstructions of Old Chinese

Several authors have produced reconstructions of liquids. Earlier systems proposed voiced final stops to account for contacts between stop-final syllables and other tones, but many investigators now believe that Old Chinese lacked tonal distinctions, with Middle Chinese tones derived from consonant clusters at the end of the syllable.


  • Sources of evidence 1
    • Middle Chinese 1.1
    • Phonetic series 1.2
    • Poetic rhyming 1.3
    • Min dialects 1.4
    • Other evidence 1.5
  • Systems 2
    • Karlgren (1940–1957) 2.1
    • Yakhontov (1959–1965) 2.2
    • Pulleyblank (1962) 2.3
    • Li (1971) 2.4
    • Baxter (1992) 2.5
    • Zhengzhang (1981–1995) 2.6
    • Baxter–Sagart (2014) 2.7
  • Comparison 3
    • Initials 3.1
    • Finals 3.2
  • Notes 4
  • References 5
  • External links 6

Sources of evidence

The major sources for the sounds of Old Chinese, covering most of the lexicon, are the sound system of Middle Chinese (7th century AD), the structure of Chinese characters, and the rhyming patterns of the Classic of Poetry (Shijing), dating from the early part of the 1st millennium BC.[1] Several other kinds of evidence are less comprehensive, but provide valuable clues. These include Min dialects, early Chinese transcriptions of foreign names, early loans between Chinese and neighbouring languages, and families of Chinese words that appear to be related.[2]

Middle Chinese

two pages of a Chinese dictionary, comprising the end of the index and the start of the entries
The start of the first rhyme class (東 dōng "east") of the Guangyun rhyme dictionary

Middle Chinese, or more precisely Early Middle Chinese, is the phonological system of the Qieyun, a rhyme dictionary published in 601, with many revisions and expansions over the following centuries. These dictionaries indicated pronunciation by dividing a syllable into an initial consonant and the rest, called the final. In his Qièyùn kǎo (1842), the Cantonese scholar Chen Li performed a systematic analysis of a later redaction of the Qieyun, identifying its initial and final categories, though not the sounds they represented. Scholars have attempted to determine the phonetic content of the various distinctions by comparing them with rhyme tables from the Song dynasty, pronunciations in modern varieties and loans in Korean, Japanese and Vietnamese (the Sinoxenic materials), but many details regarding the finals are still disputed. According to its preface, the Qieyun did not reflect a single contemporary dialect, but incorporated distinctions made in different parts of China at the time (a diasystem).[3][4]

The fact that the Qieyun system contains more distinctions than any single contemporary form of speech means that it retains additional information about the history of the language. The large number of initials and finals are unevenly distributed, suggesting hypotheses about earlier forms of Chinese.[5] For example, it includes 37 initials, but in the early 20th century Huang Kan observed that only 19 of them occurred with a wide range of finals, implying that the others were in some sense secondary developments.[6]

Phonetic series

page of a Chinese dictionary, with headings in seal script and entries in conventional script
Page from a copy of a Song dynasty edition of the Shuowen Jiezi, showing characters with the 言 element

Although the Chinese writing system is not alphabetic, the vast majority of characters are phono-semantic compounds, in which a word is written by combining a character for a similarly sounding word with a semantic indicator. Often characters sharing a phonetic element (forming a phonetic series) are still pronounced alike, as in the character 中 (zhōng, "middle"), which was adapted to write the words chōng ("pour", 沖) and zhōng ("loyal", 忠).[7] In other cases the words in a phonetic series have very different sounds both in Middle Chinese and in modern varieties. Since the sounds are assumed to have been similar at the time the characters were chosen, such relationships give clues to the lost sounds.[8]

The first systematic study of the structure of Chinese characters was Xu Shen's Shuowen Jiezi (100 AD).[9] The Shuowen was mostly based on the small seal script standardized in the Qin dynasty.[10] Earlier characters from oracle bones and Zhou bronze inscriptions often reveal relationships that were obscured in later forms.[11]

Poetic rhyming

Rhyme has been a consistent feature of Chinese poetry. While much old poetry still rhymes in modern varieties of Chinese, Chinese scholars have long noted exceptions. This was attributed to lax rhyming practice of early poets until the late-Ming dynasty scholar Chen Di argued that a former consistency had been obscured by sound change. This implied that the rhyming practice of ancient poets recorded information about their pronunciation. Scholars have studied various bodies of poetry to identify classes of rhyming words at different periods.[12][13]

The oldest such collection is the Shijing, containing songs ranging from the 10th to 7th centuries BC. The systematic study of Old Chinese rhymes began in the 17th century, when Gu Yanwu divided the rhyming words of the Shijing into ten groups (韻部 yùnbù). Gu's analysis was refined by Qing dynasty philologists, steadily increasing the number of rhyme groups. One of these scholars, Duan Yucai, stated the important principle that characters in the same phonetic series would be in the same rhyme group,[1] making it possible to assign almost all words to rhyme groups. A final revision by Wang Li in the 1930s produced the standard set of 31 rhyme groups.[15][16]

Min dialects

The Min dialects are believed to have split off before the Middle Chinese stage, because they contain distinctions that cannot be derived from the Qieyun system.

For example, the following dental initials have been identified in reconstructed proto-Min:[17]
Voiceless stops Voiced stops Nasals Laterals
Example word
Proto-Min initial *t *-t *th *d *-d *dh *n *nh *l *lh
Middle Chinese initial t th d n l

Other points of articulation show similar distinctions within stops and nasals. Proto-Min voicing is inferred from the development of Min tones, but the phonetic values of the initials are otherwise uncertain. The sounds indicated as *-t, *-d, etc. are known as "softened stops" due to their reflexes in Jianyang and nearby Min varieties in Fujian, where they appear as fricatives or approximants, or are missing entirely, while the non-softened variants appear as stops. Evidence from early loans into Mienic languages suggests that the softened stops were prenasalized.[18]

Other evidence

Several early texts contain transcriptions of foreign names and terms using Chinese characters for their phonetic values. Of particular importance are the many Buddhist transcripions of the Eastern Han period, because the native pronunciation of the source languages, such as Sanskrit and Pali, is known in detail.[19][13][20]

Eastern Han commentaries on the classics contain many remarks on the pronunciations of particular words, which has yielded a great deal of information on the pronunciations and even dialectal variation of the period.[21] By studying such glosses, the Qing philologist Qian Daxin discovered that the labio-dental and retroflex stop initials identified in the rhyme table tradition were not present in the Han period.[22][23]

Many students of Chinese have noted "word families", groups of words with related meanings and variant pronunciations, sometimes written using the same character.[24] One common case is "derivation by tone change", in which words in the departing tone appear to be derived from words in other tones.[25] Another alternation involves transitive verbs with an unvoiced initial and passive or stative verbs with a voiced initial, though scholars are divided on which form is basic.[26][27]

In the earliest period, Chinese was spoken in the valley of the Yellow River, surrounded by neighbouring languages, some of whose relatives, particularly Austroasiatic and the Tai–Kadai and Miao–Yao languages, are still spoken today. The earliest borrowings in both directions provide further evidence of Old Chinese sounds, though complicated by uncertainty about the reconstruction of early forms of those languages.[28]


Many authors have produced their own reconstructions of Old Chinese. A few of the most influential are listed here.

Karlgren (1940–1957)

The first complete reconstruction of Old Chinese was produced by the Swedish linguist Bernhard Karlgren in a dictionary of Middle and Old Chinese, the Grammata Serica (1940), revised in 1957 as the Grammata Serica Recensa. Although Karlgren's Old Chinese reconstructions have been superseded, his comprehensive dictionary remains a valuable reference for students of Old Chinese, and characters are routinely identified by their GSR position.[29] Karlgren's remained the most commonly used until it was superseded by the system of Li Fang-Kuei in the 1970s.[30]

In his Études sur la phonologie chinoise (1915–1926), Karlgren produced the first complete reconstruction of Middle Chinese (which he called "Ancient Chinese"). He presented his system as a narrow transcription of the sounds of the standard language of the Tang dynasty. Beginning with his Analytical Dictionary of Chinese and Sino-Japanese (1923), he compared these sounds across groups of words written with Chinese characters with the same phonetic component. Noting that such words were not always pronounced identically in Middle Chinese, he postulated that their initials had a common point of articulation in an earlier phase he called "Archaic Chinese", but which is now usually called Old Chinese. For example, he postulated velar consonants as initials in the series

  • kâm, 柑 kâm, 酣 ɣâm, 鉗 gjäm[31][2]

In rarer cases where different types of initials occurred in the same series, as in

  • kâk, 胳 kâk, 格 kak, 絡 lâk, 駱 lâk, 略 ljak

he postulated initial clusters *kl- and *gl-.[32]

Karlgren believed that the voiced initials of Middle Chinese were aspirated, and projected these back onto Old Chinese. He also proposed a series of unaspirated voiced initials to account for other correspondences, but later workers have discarded these in favour of alternative explanations.[33] Karlgren accepted the argument of the Qing philologist Qian Daxin that the Middle Chinese dental and retroflex stop series were not distinguished in Old Chinese, but otherwise proposed the same points of articulation in Old Chinese as in Middle Chinese.

This led him to the following series of initial consonants:[34]
Stop or Affricate Nasal Fricative/
voiceless aspirate voiced voiced aspirate voiceless voiced
Labials *p *ph *(b) *bh *m
Dentals *t *th *d *dh *n *l
Sibilants *ts *tsh *dz *dzh *s *z
Supradentals *tṣ *tṣh *dẓh *ṣ
Palatals *ťh *ďh
Velars *k *kh *g *gh *ng *x
To account for the broad variety of vowels in his reconstruction of Middle Chinese, Karlgren also proposed a complex inventory of Old Chinese vowels:[35]
*e *ộ

He also had a secondary vowel *i, which occurred only in combination with other vowels. As with Middle Chinese, Karlgren viewed his reconstruction as a narrow transcription of the sounds of Old Chinese. Thus *e rhymed with *ĕ in the Shijing, *a rhymed with *ă and *â, *ɛ rhymed with *ĕ and *ŭ, *ŭ rhymed with *u, *ô rhymed with *ộ, and *o rhymed with *ǒ and *å.[35]

Karlgren projected the final consonants of Middle Chinese, semivowels /j/ and /w/, nasals /m/, /n/ and /ŋ/, and stops /p/, /t/ and /k/ back onto Old Chinese. He also noted many cases where words in the departing tone rhymed or shared a phonetic element with words ending in a stop, e.g.

  • lâi- "depend on" and 剌 lât "wicked"[36]
  • khəi- "cough" and 刻 khək "cut, engrave"[37]

He suggested that the departing tone words in such pairs had ended with a final voiced stop (*-d or *-ɡ) in Old Chinese.[38] To account for occasional contacts between Middle Chinese finals -j and -n, Karlgren proposed that -j in such pairs derived from Old Chinese *-r.[39] He believed there was insufficient evidence to support definitive statements about Old Chinese tones.[40]

Yakhontov (1959–1965)

In a pair of papers published in 1960, the Russian linguist Sergei Yakhontov proposed two revisions to the structure of Old Chinese that are now widely accepted. He proposed that both the retroflex initials and the division-II vowels of Middle Chinese derived from the Old Chinese medial *-l- that Karlgren had proposed to account for phonetic series contacts with l-.[41] Yakhontov also observed that the Middle Chinese semi-vowel -w- had a limited distribution, occurring either after velar or laryngeal initials or before finals -aj, -an or at. He suggested that -w- had two sources, deriving from either a new series of labio-velar and labio-laryngeal initials, or from a vowel -o-, which subsequently broke to -wa- before dental codas.[42][43]

Initial consonants of Yakhontov's system[44][45]
Stop or Affricate Nasal Fricative Lateral
voiceless aspirate voiced
Labials *p *ph *bh *m
Dentals *t *th *dh *d *n *l
Sibilants *ts *tsh *dzh *s
Velars *k *kh *gh *g *ng *x
Labiovelars *kʷ *khʷ *ghʷ *gʷ *ngʷ *xʷ
Labiolaryngeals *ʔʷ
Yakhontov proposed a simpler seven-vowel system:[46][47]
Front Back
Close *e *u
Open *o

However these vowels had an uneven distribution, with *ä and *â almost in complementary distribution and *ü occurring only in open syllables and before *-k.[48] His final consonants were the nasals *-m, *-n and *-ng, corresponding stops *-p, *-t and *-k, as well as *-r, which became -j or disappeared in Middle Chinese.[49][47]

Pulleyblank (1962)

The Canadian sinologist Edwin Pulleyblank published a reconstruction of the consonants of Old Chinese in two parts in 1962. In addition to new analyses of the traditional evidence, he also made substantial use of transcription evidence. Though not a full reconstruction, Pulleyblank's work has been very influential, and many of his proposals are now widely accepted.

Pulleyblank adapted Dong Tonghe's proposal of a voiceless counterpart to the initial *m, proposing a full set of aspirated nasals,[50] as well as Yakhontov's labio-velar and labio-laryngeal initials.[51]

Pulleyblank also accepted Yakhontov's expanded role for the medial *-l-, which he noted was cognate with Tibeto-Burman *-r-.[52] To account for phonetic contacts between Middle Chinese l- and dental initials, he also proposed an aspirated lateral *lh-.[53] Pulleyblank also distinguished two sets of dental series, one derived from Old Chinese dental stops and the other derived from dental fricatives *δ and *θ, cognate with Tibeto-Burman *l-.[54] He considered recasting his Old Chinese *l and *δ as *r and *l to match the Tibeto-Burman cognates, but rejected the idea to avoid complicating his account of the evolution of Chinese.[55] Later he re-visited this decision, recasting *δ, *θ, *l and *lh as *l, *hl, *r and *hr respectively.[56]

Pulleyblank also proposed an Old Chinese labial fricative *v for the few words where Karlgren had *b, as well as a voiceless counterpart *f.[57] Unlike the above ideas, these have not been adopted by later workers.

Initial consonants of Pulleyblank's system[58]
Stop or Affricate Nasal Fricative Lateral
voiceless aspirate voiced aspirate voiced voiceless voiced aspirate voiced
Labials *p *ph *b *mh *m *f *v
Dentals *t *th *d *nh *n *lh *l
Sibilants *ts *tsh *dz *s
Velars *k *kh *g *ŋh
Labiovelars *kw *khw *gw *ŋhw *ŋw
Laryngeals[3] *h
Labiolaryngeals *ʔw *hw *ɦw

Pulleyblank also proposed a number of initial consonant clusters, allowing any initial to be preceded by *s- and followed by *-l- (*-r- in later revisions), and grave initials and *n to be followed by *-δ- (*-l- in later revisions).[58]

On the basis of transcription evidence, Pulleyblank argued that the -j- medial of Middle Chinese was an innovation not present in Old Chinese. He classified Middle Chinese finals without -j- as type A and those with the medial as type B, and suggested that they arose from Old Chinese short and long vowels respectively.[59]

Vietnamese were derived from final consonants *-ʔ and *-s in an atonal ancestral language. He also suggested that the Chinese departing tone derived from earlier *-s. Then the departing tone syllables that Karlgren had reconstructed with *-d and *-g could instead be reconstructed as *-ts and *-ks, with the stops subsequently being lost before the final *-s, which eventually became a tonal distinction. The absence of a corresponding labial final could be attributed to early assimilation of *-ps to *-ts. Pulleyblank strengthened the theory with several examples of syllables in the departing tone being used to transcribe foreign words ending in -s into Chinese, and also suggested that *-s acted as a derivational suffix in Old Chinese.[60] He further proposed that the Middle Chinese rising tone derived from *-ʔ, implying that Old Chinese lacked tones.[61] Mei Tsu-lin later supported this theory with evidence from early transcriptions of Sanskrit words, and pointed out that rising tone words end in a glottal stop in some modern Chinese dialects, including Wenzhounese and some Min dialects.[62]

Li (1971)

The Chinese linguist Li Fang-Kuei published an important new reconstruction in 1971, synthesizing proposals of Yakhontov and Pulleyblank with ideas of his own. His system remained the most commonly used until it was replaced by that of Baxter in the 1990s. Although Li did not produce a complete dictionary of Old Chinese, he presented his methods in sufficient detail that others could apply them to the data.[63] Schuessler (1987) includes reconstructions of the Western Zhou lexicon using Li's system.[30]

Li included the labio-velars, labio-laryngeals and voiceless nasals proposed by Pulleyblank. As Middle Chinese g- occurs only in palatal environments, Li attempted to derive both g- and ɣ- from Old Chinese *g- (and similarly *gw-), but had to assume irregular developments in some cases.[64][65]

Thus he arrived at the following inventory of initial consonants:[66][67]
Stop or Affricate Nasal Lateral Fricative/
voiceless aspirate voiced voiceless voiced voiceless voiced voiceless voiced
Labials *p *ph *b *hm *m
Dentals *t *th *d *hn *n *hl *l *r
Sibilants *ts *tsh *dz *s
Velars *k *kh *g *hng *ng
Labiovelars *kw *khw *gw *hngw *ngw
Laryngeals[4] *h
Labiolaryngeals *ʔw *hw

Li also included the *-l- medial proposed by Pulleyblank, in most cases re-interpreting it as *-r-. In addition to the medial *-j- projected back from Middle Chinese, he also postulated the combination *-rj-.[68]

Assuming that rhyming syllables had the same main vowel, Li proposed a system of four vowels *i, *u, and *a.

He also included three diphthongs *iə, *ia and *ua to account for syllables that were placed in rhyme groups reconstructed with or *a but were distinguished in Middle Chinese:[69]
*i *u
*ia *a *ua

Li followed Karlgren in proposing finals consonants *-d and *-g, but was unable to clearly separate them from open syllables, and extended them the all rhyme groups but one, for which he proposed a final *-r.[70] He also proposed that labio-velar consonants could occur as final consonants.

Thus in Li's system every syllable ended in one of the following consonants:[71]
*p *m
*r *d *t *n
*g *k *ng
*gw *kw *ngw

Li marked the rising and departing tones with a suffix *-x or *-h, without specifying how they were realized.[72]

Baxter (1992)

William H. Baxter's monograph A Handbook of Old Chinese Phonology displaced Li's reconstruction in the 1990s. Baxter did not produce a dictionary of reconstructions, but the book contains a large number of examples, including all the words occurring in rhymes in the Shijing, and his methods are described in great detail. Schuessler (2007) contains reconstructions of the entire Old Chinese lexicon using a simplified version of Baxter's system.

Baxter's treatment of the initials is largely similar to the proposals of Pulleyblank and Li. He reconstructed the liquids *l, *hl, *r and *hr in the same contexts as Pulleyblank.[73] Unlike Li, he distinguished Old Chinese *ɦ and *w from *g and *gʷ.[65]

Other additions were *z, with a limited distribution,[74] and voiceless and voiced palatals *hj and *j, which he described as "especially tentative, being based largely on scanty graphic evidence".[75]
Initial consonants of Baxter's system[76]
Stop or Affricate Nasal Lateral Fricative/
voiceless aspirate voiced voiceless voiced voiceless voiced voiceless voiced
Labials *p *ph *b *hm *m
Dentals *t *th *d *hn *n *hl *l *hr *r
Sibilants *ts *tsh *dz *s *z
Palatals *hj *j
Velars *k *kh *g *hng *ng
Labiovelars *kʷ *kʷh *gʷ *hngʷ *ngʷ
Laryngeals *x
Labiolaryngeals *ʔʷ *hw *w

As in Pulleyblank and Li's systems, the possible medials were *-r-, *-j- and the combination *-rj-.[77] However while Li had proposed *-rj- as conditioning palatalization of velars, Baxter followed Pulleyblank in proposing it as the source of division III chóngniǔ finals.[78]

Baxter's major contribution concerned the vowel system and rhyme groups. Nicholas Bodman had proposed a six-vowel system for a proto-Chinese phase, based on comparison with other Sino-Tibetan languages.[79] Baxter argued for a six-vowel system in Old Chinese by re-analysing of the traditional rhyme groups. For example, the traditional 元 rhyme group of the Shijing corresponds to three different finals in Middle Chinese. While Li had sought to reconcile these distinct outcomes from rhyming words by reconstructing the finals as *-ian, *-an and *-uan, Baxter argued that in fact they did not rhyme in the Shijing, and could thus be reconstructed with three distinct vowels, *e, *a and *o. Baxter proposed that the traditional 31 rhyme groups should be refined to over 50, and performed a statistical analysis of the actual rhymes of the Shijing, which supported the new groups with varying degrees of confidence.[80]

Baxter's six vowels[81]
*i *u
*e *a *o

Zhengzhang Shangfang and Sergei Starostin independently developed similar vowel systems.[82][83]

Baxter's final consonants were those of Middle Chinese, plus *-wk (an allophone of *-kʷ), optionally followed by a post-coda *-ʔ or *-s.[84]
MC vocalic coda MC stop coda MC nasal coda
*-p *-m *-mʔ *-ms
*-j *-jʔ *-js *-ts *-t *-n *-nʔ *-ns
*-∅ *-ʔ *-s *-ks *-k *-ŋ *-ŋʔ *-ŋs
*-w *-wʔ *-ws *-wks *-wk

Baxter also speculated on the possibility of a glottal stop occurring after oral stop finals. The evidence is limited, and consists mainly of contacts between rising tone syllables and -k finals, which could alternatively be explained as phonetic similarity.[85]

Zhengzhang (1981–1995)

Zhengzhang Shangfang (郑张尚芳) published his ideas in a series of articles in Chinese provincial journals, and were not widely disseminated. Some of his notes were translated into English by Laurent Sagart in 2000.[86]

Zhengzhang's reconstruction incorporates a suggestion by Pan Wuyun that the three Middle Chinese laryngeal initials are reflexes of uvular stops in Old Chinese, and thus parallel to the other sets of stops.[87]

He argues that Old Chinese lacked affricate initials, and that the Middle Chinese affricates reflect Old Chinese clusters of *s- and other consonants, yielding the following inventory of initial consonants:[88]
Stop or Affricate Resonant Fricative Approximant
voiceless aspirate voiced aspirate voiced voiceless voiced
Labials *p *ph *b *mh *m *w
Dentals *t *th *d *nh *n *s
*rh *r
*lh *l *j
Velars *k *kh *g *ŋh
Uvulars *q > *ʔ *qh > *h *ɢ > *ɦ

Zhengzhang's *w medial could occur only after velar and uvular initials, matching the labio-velar and labio-laryngeal initials of other reconstructions.[89]

Zhengzhang also refined the traditional rhyme classes to obtain a six-vowel system similar to those of Baxter and Starostin, but with *ɯ corresponding to Baxter's *ɨ and Starostin's *ə:[90]

*i *u
*e *a *o

Zhengzhang argued that the final stops of Old Chinese were voiced, like those of Old Tibetan.[91] He accepted the consonantal origin of Middle Chinese tones.[92]

Baxter–Sagart (2014)

Jerry Norman concluded his review of Baxter (1992) with the words:

A reader of Baxter's book is left with the impression that he has pushed the traditional approach to its limits and that any further progress in the field will have to be based on a quite different methodological approach.[93]

Sagart (1999) used Baxter's system, with minor variations, in a study of the derivational morphology of Old Chinese. Thereafter Baxter and Sagart collaborated on a revision of the reconstruction, using evidence from morphology and early loans to other languages. Preliminary materials were placed on a website in February 2011, and a monograph appeared in 2014.[94]

The initial consonants of the revised system largely correspond to those of Baxter (1992) apart from the dropping of the rare initials *z, *j and *hj. Instead of marking type B syllables with a *-j- medial, they treated type A syllables as having pharyngealized initials, adapting a proposal of Jerry Norman and thus doubling the number of initials.[95][5] They also adopted the proposal of Pan Wuyun to recast the laryngeal initials as uvular stops, though they retained a separate glottal stop.[97]

Initials of the Baxter-Sagart system
Stop or Affricate Nasal Lateral Fricative/
voiceless aspirate voiced voiceless voiced voiceless voiced voiceless voiced
Labials *p, *pˁ *pʰ, *pʰˁ *b, *bˁ *m̥, *m̥ˁ *m, *mˁ
Dentals *t, *tˁ *tʰ, *tʰˁ *d, *dˁ *n̥, *n̥ˁ *n, *nˁ *l̥, *l̥ˁ *l, *lˁ *r̥, *r̥ˁ *r, *rˁ
Sibilants *ts, *tsˁ *tsʰ, *tsʰˁ *dz, *dzˁ *s, *sˁ
Velars *k, *kˁ *kʰ, *kʰˁ , *ɡˁ *ŋ̊, *ŋ̊ˁ , *ŋˁ
Labiovelars *kʷ, *kʷˁ *kʷʰ, *kʷʰˁ *ɡʷ, *ɡʷˁ *ŋ̊ʷ, *ŋ̊ʷˁ *ŋʷ, *ŋʷˁ
Uvulars *q, *qˁ *qʰ, *qʰˁ , *ɢˁ
Labio-uvulars *qʷ, *qʷˁ *qʰʷ, *qʰʷˁ *ɢʷ, *ɢʷˁ
Laryngeals , *ʔˁ
Labiolaryngeals (*ʔʷˁ)
They propose the uvular initial as a second source of the Middle Chinese palatal initial in addition to *l, so that series linking Middle Chinese y- with velars or laryngeals instead of dentals are reconstructed as uvulars rather than laterals, for example[97]
Old Chinese
Baxter (1992) Baxter–Sagart
kjoX *kljaʔ *C.qaʔ
yoX *ljaʔ *C.ɢaʔ

Baxter and Sagart retained the six-vowel system, though re-casting *ɨ as *ə. The finals consonants were unchanged except for the addition of a final *r in syllables showing connections between final consonants -j and -n in Middle Chinese, as suggested by Sergei Starostin.[98]

The major departure from Baxter's system lay in the structure of roots proposed by Sagart, in which roots could comprise either a monosyllable or a syllable preceded by a preinitial consonant, in one of two patterns:[99]

  • a "tightly attached" preinitial forming a consonant cluster, as in 肉 *k.nuk "flesh", 用 *m.loŋ-s "use" and 四 *s.lij-s "four", and
  • a "loosely attached" preinitial, forming a minor syllable, as in 脰 *kə.dˤok-s "neck", 舌 *mə.lat "tongue" and 脣 *sə.dur "lip".

Similar root structures are found in the modern rGyalrong, Khmer and Atayal languages.[100] Sagart argued that such iambic combinations, like single syllables, were written with single characters and also counted as a single foot in verse.[101] The various initials are reconstructed based on comparisons with proto-Min cognates and early loans to Hmong–Mien languages and Vietnamese:[102]

Reconstructed initials, illustrated with labials
Middle Chinese proto-Min proto-Hmong–Mien Vietnamese Baxter–Sagart Old Chinese
ph *ph *pʰ ph H *pʰ(ˤ)[103]
*mp ? *mə.pʰ(ˤ)[104]
p *p *p b H *p(ˤ)[105]
? v H *C.p(ˤ)[106]
*-p *mp ? *mə.p(ˤ), *Nə.p(ˤ)[107]
b *bh ? v H,L *m.p(ˤ)[108]
*mb ? *m.b(ˤ), *C.b(ˤ)[109]
*b *b b L *b(ˤ)[110]
? b H *N.p(ˤ)[111]
*-b *mb v L *mə.b(ˤ), *Cə.b(ˤ)[112]
m *mh *hm m H *C.m(ˤ)[113]
*m *m m L *m(ˤ)[114]
x(w) *x ? ? *m̥(ˤ)[115]


The different reconstructions provide different interpretations of the relationships between the categories of Middle Chinese and the main bodies of ancient evidence: the phonetic series (used to reconstruct initials), and the Shijing rhyme groups (used to reconstruct finals).


Karlgren first stated the principle that words written with the same phonetic component had initials with a common point of articulation in Old Chinese. Moreover, nasal initials seldom interchanged with other consonants.[116] Thus phonetic series can be placed into classes, depending on the range of Middle Chinese initials found in them, and these classes are presumed to correspond to classes of Old Chinese initials.[117] Where markedly different Middle Chinese initials occur together in a series, investigators have proposed additional consonants, or clusters of consonants, in Old Chinese.

Reconstructions of Old Chinese initials in various types of phonetic series
Type of series Middle Chinese Examples Old Chinese reconstructions
Li Baxter Karlgren Pulleyblank Li Baxter
Labial stops[118] p- p- *p- *p- *p- *p-
ph- ph- *p'- *ph- *ph- *ph-
b- b- *b'- *b- *b- *b-
l- l- *bl- *vl- *bl- *b-r-
Labial nasal[119] m- m- *m- *m- *m- *m-
x(w)- x(w)- *xm- *mh- *hm- *hm-
Dental stops,
retroflex stops
and palatals[120]
t- t- *t- *t- *t- *t-
th- th- *t'- *th- *th- *th-
d- d- *d'- *d- *d- *d-
ṭ- tr- *t- *tl- *tr- *tr-
ṭh- trh- *t'- *thl- *thr- *thr-
ḍ- dr- 𣥺 *d- *dl- *dr- *dr-
tś- tsy- *ťi̯- *t-[6] *tj- *tj-
tśh- tsyh- *ť'i̯- *th-[6] *thj- *thj-
dź- dzy- *ďi̯- *d-[6] *dj- *dj-
ś- sy- *śi̯- ? *sthj- *stj-
Dental stops,
s- and j-[121]
th- th- *t'- *θ- *th- *hl-
d- d- *d'- *δ- *d- *l-
ṭh- trh- *t'- *θl- *thr- *hlr-
ḍ- dr- *di̯- *δl- *dr- *lr-
s- s- *s- *sθ- *st- *sl-
z- z- *dzi̯- *sδy- *rj- *zl-
ś- sy- *śi̯- *θ-[6] *sthj- *hlj-
ź- zy- *d'i̯- *δ-[6] *dj- *Lj-
ji- y- *di̯-, *zi̯- *δ-[6] *r- *lj-
Dental stops and l-[122] l- l- *l- *l- *l- *C-r-
th- th- *t'l- *lh- *hl- *hr-
ṭh- trh- *t'l- *lh- *hlj- *hrj-
Dental nasal[123] n- n- *n- *n- *n- *n-
ṇ- nr- *ni̯- *nl- *nr- *nr-
ńź- ny- *ńi̯- *nj- *nj- *nj-
th- th- *t'n- *nh- *hn- *hn-
ṭh- trh- *t'n- *nhl- *hnr- *hnr-
ś- sy- *śńi̯- *nh-[6] *hnj- *hnj-
s- s- *sni̯- *snh- *sn- *sn-
Sibilants[124] ts- ts- *ts- *ts- *ts- *ts-
tsh- tsh- *ts'- *tsh- *tsh- *tsh-
dz- dz- *dz'- *dz- *dz- *dz-
s- s- *s- *s- *s- *s-
tṣ- tsr- *tṣ- *tsl- *tsr- *tsr-
tṣh- tsrh- *tṣ'- *tshl- *tshr- *tshr-
dẓ- dzr- *dẓ'- *dzl- *dzr- *dzr-
ṣ- sr- *s- *sl- *sr- *srj-
Velars and palatals[125] k- k- *k- *k- *k- *k-
kh- kh- *k'- *kh- *kh- *kh-
ɣ- h- *g'- *g- *g- *g-
g- g- *g'i̯- *gy- *gj- *gj-
tś- tsy- *ťi̯- *ky- *krj- *kj-
tśh- tsyh- *ť'i̯- *khy- *khrj- *khj-
dź- dzy- *ďi̯- *gy- *grj- *gj-
l- l- *gl- *ɦl- *gl- *g-r-
Laryngeals[126] ʔ- ʔ- *ʔ- *ʔ- *ʔ- *ʔ-
x- x- *x- *x- *x- *x-
ɣ- h- *g'- *ɦ- *g- *ɦ-
j- hj- *gi̯- *ɦ-[6] *gj- *ɦj-
Velar nasal[127] ng- ng- *ng- *ŋ- *ng- *ng-
x- x- *x- *ŋh- *hng- *hng-
ńź- ny- *ńi̯- *ŋy- *ngrj- *ngj-
ś- sy- *śńi̯- *ŋhy- *hngrj- *hngj-
Labio-velars[128] kw- kw- *kw- *kw- *kw- *kʷ-
khw- khw- *k'w- *khw- *khw- *kʷh-
ɣw- hw- *g'w- *gw- *gw- *gʷ-
gw- gw- *g'wi̯- gwy- *gwj- *gʷj-
Labio-laryngeals[129] ʔw- ʔw- *ʔw- *ʔw- *ʔw- *ʔʷ-
xw- xw- *xw- *xw- *xw- *hw-
ɣw- hw- *g'w- *ɦw- *gw- *w-
jw- hwj- *gi̯w- *ɦw-[6] *gwj- *wj-
Labio-velar nasal[130] ngw- ngw- *ngw- *ŋw- *ngw- *ngʷ-
xw- xw- *xw- *ŋhw- *hngw- *hngʷ-


Karlgren noted that the finals of Middle Chinese can be divided into a number of classes, which combine with different groups of initials. These distributional classes are partially aligned with the placement of finals in different rows of the Song dynasty rhyme tables. As three classes of final occurred in the first, second and fourth rows respectively, he named them finals of divisions I, II and IV. The remaining finals he called "division-III finals" because they occurred in the third row of the tables. Some of these (the "pure" or "independent" division-III finals) occurred only in that row, while others (the "mixed" finals) could also occur in the second or fourth rows with some initials.[131] Karlgren disregarded the chongniu distinction, but later workers have emphasized its importance. Li Rong, in a systematic comparison of the rhyme tables with a recently discovered early edition of the Qieyun, identified seven classes of finals. The table below lists the combinations of initial and final classes that occur in the Qieyun, with the row of the rime tables in which each combination was placed:[132][133]

Classes of Middle Chinese finals, with rows in the rhyme tables
Phonetic series Middle Chinese initials Middle Chinese finals
div. I div. II division-III div. IV
pure mixed chongniu
Labials Labials 1 2 3 3 3 4 4
Dentals Dental stops 1 4
Retroflex stops 2 3 3
Dentals, velars Palatal sibilants 3 3
Sibilants Dental sibilants 1 4 4 4
Retroflex sibilants 2 2 2
Velars Velars 1 2 3 3 3 4 4
Laryngeals Laryngeals 1 2 3 3 3 4 4

Of these classes, the division I and IV finals are believed to be primary, while those of other classes were modified by medials.[134] These two classes have identical distributions in the rhyme dictionaries.[135]

Qing scholars had studied the rhyming patterns of the Shijing, classifying words into rhyme groups, refined to 31 groups by the early 20th century.[15] Parallels had also been noted between rhyme groups of three different types: those with a vocalic coda in Middle Chinese (the 陰 yīn groups), those with a stop coda (the 入 or entering tone), and those with a corresponding nasal coda (the 陽 yáng groups).[71][136] Most workers assume that words that rhymed in the Shijing had the same main vowel and the same final consonant, though they differ on the particular vowels reconstructed. The six-vowel systems of Baxter, Starostin and Zhengzhang imply that the traditional rhyme groups must be split more finely.[137]

Reconstructions of Old Chinese vowels in division I and IV parts of Shijing rhyme groups[138][139]
Shijing rhyme groups and corresponding Middle Chinese finals in Li's and Baxter's notations Reconstructed Old Chinese vowels
MC vocalic coda
MC stop coda
MC nasal coda
Karlgren Li Baxter Zhengzhang
-iep / -ep -iem / -em *iə *i *i
-əp / -op -əm / -om *ɨ, *u *ɯ, *u
/ -iep / -ep -iem / -em *ia *ia *e *e
-âp / -ap -âm / -am *â, *a, *ă *a *a, *o *a, *o
-iei / -ej -iet / -et -ien / -en *ĕ, *e *i *i *i
-iei / -ej / -iet / -et / -ien / -en *iə
-əi / -oj -ət / -ot -ən / -on *u *u
[7] -iei / -ej -iet / -et -ien / -en *ia *ia *e *e
-â / -a -âi / -aj -ât / -at -ân / -an *â, *a, *ă *a *a *a
-wâ / -wa -wâi / -waj -wât / -wat -wân / -wan *wâ, *wa, *wă *ua *o *o
/ -iei / -ej -iek / -ek -ieng / -eng *ĕ, *e *i *e *e
-əi / -oj -ək / -ok -əng / -ong *ə, *ɛ, *ŭ
-uo / -u -âk / -ak -âng / -ang *â, *a, *ă *a *a *a
-əu / -uw -uk / -uwk -ung / -uwng *u, *ŭ *u *o *o
-âu / -aw / -uok / -owk / -uong / -owng *ô, *ộ *ə-ʷ *u *u
-ieu / -ew -iek / -ek *iô, *iộ *iə-ʷ *i-ʷ *i-ʷ, *ɯ-ʷ
-âu / -aw -âk / -ak *o, *ǒ, *å *a-ʷ *a-ʷ *a-ʷ, *o-ʷ
-ieu / -ew -iek / -ek *io *ia-ʷ *e-ʷ *e-ʷ
: Old Chinese finals reconstructed with labiovelar codas


  1. ^ 同聲必同部 Tóng shēng bì tóng bù.[14]
  2. ^ Middle Chinese forms given in Li's revision of Karlgren's notation.
  3. ^ Pulleyblank wrote the glottal stop as "·".[58]
  4. ^ Li wrote the glottal stop as "·".[66]
  5. ^ Norman originally proposed pharyngealization only in type A syllables without retroflexion (the *-r- medial).[96]
  6. ^ a b c d e f g h i followed by a long vowel, causing palatalization.
  7. ^ The 祭 group included departing tone words only.[140]


  1. ^ Baxter (1992), pp. 2–3.
  2. ^ Baxter (1992), pp. 12–13, 25.
  3. ^ Baxter (1992), pp. 32–44.
  4. ^ Norman (1988), pp. 24–42.
  5. ^ Baxter (1992), pp. 37–38.
  6. ^ Zhengzhang (2000), pp. 12–13.
  7. ^ GSR 1007a,p,k.
  8. ^ Norman (1988), pp. 43–44.
  9. ^ Baxter (1992), p. 13.
  10. ^ Baxter (1992), pp. 346–347.
  11. ^ Baxter (1992), p. 5.
  12. ^ Norman (1988), p. 42.
  13. ^ a b Baxter (1992), p. 12.
  14. ^ Baxter (1992), p. 831.
  15. ^ a b Baxter (1992), pp. 150–170.
  16. ^ Norman (1988), pp. 42–44.
  17. ^ Norman (1973); Norman (1988), pp. 228–229.
  18. ^ Norman (1986), p. 381.
  19. ^ Pulleyblank (1992), p. 375–379.
  20. ^ Coblin (1983), p. 7.
  21. ^ Coblin (1983), pp. 4–7.
  22. ^ Norman (1988), p. 44.
  23. ^ Dong (2014), pp. 33–35.
  24. ^ Pulleyblank (1973).
  25. ^ Downer (1959).
  26. ^ Schuessler (2007), p. 49.
  27. ^ Handel (2012), pp. 63–71.
  28. ^ Sagart (1999), pp. 8–9.
  29. ^ Handel (2003), p. 547.
  30. ^ a b Handel (2003), p. 548.
  31. ^ Karlgren (1923), p. 18.
  32. ^ Karlgren (1923), p. 31.
  33. ^ Li (1974–75), pp. 230–231.
  34. ^ Li (1974–75), p. 230.
  35. ^ a b Li (1974–75), p. 244.
  36. ^ GSR 272e,a.
  37. ^ GSR 937s,v.
  38. ^ Karlgren (1923), pp. 27–30.
  39. ^ Baxter (1992), p. 843.
  40. ^ Karlgren (1957), p. 2.
  41. ^ Baxter (1992), pp. 23, 178, 262.
  42. ^ Baxter (1992), pp. 180, 250.
  43. ^ Yakhontov (1970).
  44. ^ Yakhontov (1965), p. 30.
  45. ^ Yakhontov (1978–79), p. 39.
  46. ^ Yakhontov (1965), p. 27.
  47. ^ a b Yakhontov (1978–79), p. 37.
  48. ^ Yakhontov (1978–79), p. 38.
  49. ^ Yakhontov (1965), p. 26.
  50. ^ Pulleyblank (1962a), pp. 92–93, 121, 135–137.
  51. ^ Pulleyblank (1962a), pp. 95–96.
  52. ^ Pulleyblank (1962a), pp. 110–114.
  53. ^ Pulleyblank (1962a), pp. 121–122.
  54. ^ Pulleyblank (1962a), pp. 114–119.
  55. ^ Pulleyblank (1962a), p. 117.
  56. ^ Pulleyblank (1973), p. 117.
  57. ^ Pulleyblank (1962a), pp. 137–141.
  58. ^ a b c Pulleyblank (1962a), p. 141.
  59. ^ Pulleyblank (1962a), pp. 141–142.
  60. ^ Norman (1988), pp. 54–57.
  61. ^ Pulleyblank (1962b), pp. 209–235.
  62. ^ Mei (1970).
  63. ^ Norman (1988), p. 45.
  64. ^ Li (1974–75), p. 235.
  65. ^ a b Baxter (1992), pp. 209–210.
  66. ^ a b Li (1974–75), p. 237.
  67. ^ Norman (1988), p. 46.
  68. ^ Li (1974–75), pp. 237–240.
  69. ^ Li (1974–75), pp. 243–247.
  70. ^ Baxter (1992), pp. 331–333.
  71. ^ a b Norman (1988), p. 48.
  72. ^ Li (1974–75), pp. 248–250.
  73. ^ Baxter (1992), pp. 196–202.
  74. ^ Baxter (1992), p. 206.
  75. ^ Baxter (1992), pp. 202–203.
  76. ^ Baxter (1992), p. 177.
  77. ^ Baxter (1992), pp. 178–180.
  78. ^ Baxter (1992), pp. 178–179, 214.
  79. ^ Bodman (1980), p. 47.
  80. ^ Baxter (1992), pp. 367–564.
  81. ^ Baxter (1992), p. 180.
  82. ^ Zhengzhang (2000), pp. 42–43.
  83. ^ Starostin (1989), pp. 343–429.
  84. ^ Baxter (1992), pp. 181–183.
  85. ^ Baxter (1992), pp. 323–324.
  86. ^ Zhengzhang (2000), p. vii.
  87. ^ Zhengzhang (2000), p. 14.
  88. ^ Zhengzhang (2000), p. 18.
  89. ^ Zhengzhang (2000), p. 25.
  90. ^ Zhengzhang (2000), pp. 33–43.
  91. ^ Zhengzhang (2000), pp. 60–61.
  92. ^ Zhengzhang (2000), pp. 63–68.
  93. ^ Norman (1993), p. 705.
  94. ^ Baxter & Sagart (2014).
  95. ^ Baxter & Sagart (2014), pp. 43, 68–76.
  96. ^ Norman (1994).
  97. ^ a b Sagart (2007).
  98. ^ Baxter & Sagart (2014), pp. 252–268.
  99. ^ Sagart (1999), pp. 14–15.
  100. ^ Sagart (1999), p. 13.
  101. ^ Baxter & Sagart (2014), p. 53.
  102. ^ Baxter & Sagart (2014), pp. 83–99.
  103. ^ Baxter & Sagart (2014), pp. 102–105.
  104. ^ Baxter & Sagart (2014), pp. 177–178.
  105. ^ Baxter & Sagart (2014), pp. 99–102.
  106. ^ Baxter & Sagart (2014), pp. 168–169.
  107. ^ Baxter & Sagart (2014), pp. 174, 176–177.
  108. ^ Baxter & Sagart (2014), pp. 123–127.
  109. ^ Baxter & Sagart (2014), pp. 131–134, 170–172.
  110. ^ Baxter & Sagart (2014), pp. 105–108.
  111. ^ Baxter & Sagart (2014), pp. 116–119.
  112. ^ Baxter & Sagart (2014), pp. 178, 188–189.
  113. ^ Baxter & Sagart (2014), pp. 171–173.
  114. ^ Baxter & Sagart (2014), pp. 108–111.
  115. ^ Baxter & Sagart (2014), pp. 99, 111–112.
  116. ^ Karlgren (1923), pp. 17–18.
  117. ^ Branner (2011), pp. 132–137.
  118. ^ GSR 25, 668, 740; Pulleyblank (1962a), pp. 134–135; Baxter (1992), pp. 188, 199.
  119. ^ GSR 503, 742, 947; Pulleyblank (1962a), pp. 135–137; Li (1974–75), pp. 235–236; Baxter (1992), pp. 188–189.
  120. ^ GSR 45, 725, 961; Pulleyblank (1962a), pp. 107–109; Li (1974–75), pp. 228–229, 232–233, 242; Baxter (1992), pp. 191–195, 229.
  121. ^ GSR 82, 465, 976; Pulleyblank (1962a), pp. 114–119; Li (1974–75), pp. 231–232; Baxter (1992), pp. 196–199, 225–226.
  122. ^ GSR 23, 272, 597; Pulleyblank (1962a), pp. 121–122; Li (1974–75), p. 237; Baxter (1992), pp. 199–202.
  123. ^ GSR 94, 354, 1076; Pulleyblank (1962a), pp. 119–121, 131–133; Li (1974–75), p. 236; Baxter (1992), pp. 191–196, 222.
  124. ^ GSR 5, 798, 812; Pulleyblank (1962a), pp. 126–129; Li (1974–75), p. 232; Baxter (1992), pp. 203–206.
  125. ^ GSR 74, 552, 609; Pulleyblank (1962a), pp. 86–88, 98–107, 110–114; Li (1974–75), pp. 233, 235; Baxter (1992), pp. 199, 206–208, 210–214.
  126. ^ GSR 55, 200, 653; Pulleyblank (1962a), pp. 86–92; Li (1974–75), p. 233; Baxter (1992), pp. 207, 209–210.
  127. ^ GSR 2, 873, 1164; Pulleyblank (1962a), pp. 92–95; Li (1974–75), pp. 235–237; Baxter (1992), pp. 208–209.
  128. ^ GSR 302, 538, 992; Pulleyblank (1962a), pp. 95–98; Li (1974–75), pp. 233–235; Baxter (1992), pp. 214–216.
  129. ^ GSR 97, 255, 995; Pulleyblank (1962a), p. 95–98; Li (1974–75), pp. 233–235; Baxter (1992), pp. 217–218.
  130. ^ GSR 19; Pulleyblank (1962a), p. 92; Li (1974–75), pp. 235–237; Baxter (1992), pp. 216–217.
  131. ^ Branner (2006), p. 24.
  132. ^ Branner (2006), p. 25.
  133. ^ Baxter (1992), pp. 63–81.
  134. ^ Baxter (1992), pp. 236–258.
  135. ^ Branner (2006), pp. 32–33.
  136. ^ Pulleyblank (1977–78), p. 181.
  137. ^ Baxter (1992), pp. 180, 253–254, 813.
  138. ^ Baxter (1992), pp. 141–150, 170, 243–246, 254–255, 298–302.
  139. ^ Li (1974–75), pp. 252–279.
  140. ^ Baxter (1992), p. 389.
Works cited
  • translation of
  • translation of Chapter 2 (Phonetics) of Yakhontov (1965).

External links

  • StarLing database, by Georgiy Starostin.
  • The Baxter-Sagart reconstruction of Old Chinese.
  • 上古音查询 ("Ancient sound query", in Chinese), by Pan Wuyun.