Scholars have attempted to reconstruct thephonology ofOld Chinese from documentary evidence. Although the writing system does not describe sounds directly, sharedphonetic components of the most ancientChinese characters are believed to link words that were pronounced similarly at that time. The oldest surviving Chinese verse, in theClassic of Poetry (Shijing), shows which words rhymed in that period. Scholars have compared these bodies of contemporary evidence with the much laterMiddle Chinese reading pronunciations listed in theQieyunrhyme dictionary published in 601 AD, though this falls short of a phonemic analysis. Supplementary evidence has been drawn from cognates in otherSino-Tibetan languages and inMin Chinese, which split off before the Middle Chinese period, Chinese transcriptions of foreign names, and early borrowings from and by neighbouring languages such asHmong–Mien,Kra-Dai,Vietic andTocharian languages.

Although many details are disputed, most recent reconstructions agree on the basic structure. It is generally agreed that Old Chinese differed from Middle Chinese in lacking retroflex and palatalobstruents but having initial consonant clusters of some sort, and in having voicelesssonorants. Most recent reconstructions also posit consonant clusters at the end of the syllable, developing intotone distinctions in Middle Chinese.

The reconstruction of Old Chinese typically starts from "Early Middle Chinese", the phonological system of theQieyun, arhyme dictionary published in 601, with many revisions and expansions over the following centuries.According to its preface, theQieyun did not record a single contemporary dialect, but set out to codify the pronunciations of characters to be used when reading theclassics, incorporating distinctions made in different parts of China at the time (adiasystem).^[1]

TheQieyun and its successors grouped characters by tone class, rhyme group and homophone group.The pronunciation of each group of homophonous characters was indicated using thefanqie method, using a pair of other words with the same initial consonant and final (the rest of the syllable) respectively.^[2]Analysis of thefanqie spellings allows one to enumerate the initials and finals of the system, but not to determine their phonetic values.^[3]

Therhyme tables from theSong dynasty classify the syllables of theQieyun using a sophisticatedfeature analysis of the initials and finals, though influenced by the different pronunciations of that later period.^[4][5]Scholars have attempted to determine the phonetic content of the initials and finals by comparing them with the rhyme tables and by examining pronunciations in modern varieties and loans in Japanese, Korean and Vietnamese (theSinoxenic materials), but many details regarding the finals are still uncertain.^[6][7]

Each character of the script represented a single Old Chinesemorpheme, originally identical to a word.^[8]The system does not use symbols for individual sounds as is done in an alphabetic system.^[9]However, while a minority of words are represented pictorially, most are written by borrowing characters for similar-sounding words.Often a disambiguatingsemantic indicator element was added to the phonetic element, yielding aphono-semantic compound.^[10]Often words written with a shared phonetic element are still pronounced alike, as in the wordzhōng ('middle',中), whose character was adapted to write the wordschōng ('pour',沖) andzhōng ('loyal',忠).^[11]In other cases words sharing a phonetic element have very different sounds in any known variety of Chinese, but are assumed to have been similar in sound at the time the characters were chosen.^[12]

The first systematic study of the structure of Chinese characters was theShuowen Jiezi (100 AD) written byXu Shen.^[13]TheShuowen was mostly based on thesmall seal script standardized in theQin dynasty.^[14]Earlier characters fromoracle bones andZhou bronze inscriptions often reveal relationships that were obscured in later forms.^[15]However, these brief and formulaic texts used only a small part of the lexicon.Since the late 20th century, several caches ofbamboo and wooden slips from the Warring States period have been recovered, especially from the states ofChu andQin.^[16]These texts feature considerable variation in the characters used to represent a word.The study of these variants can clarify both the pronunciations of individual words and also which phonetic elements were considered interchangeable.^[17]

The other large body of contemporaneous evidence is rhyme in ancient texts. Most work has focussed on theShijing, a collection of songs and poetry from the 11th to 7th centuries BC.^[18]Again some of these songs still rhyme in modern varieties of Chinese, but many do not.For many centuries, this was attributed to lax rhyming practice or varied pronunciations.^[19]The puzzle was solved by the late-Ming dynasty scholarChen Di, who demonstrated that rhyming practice was consistent within theShijing. Observing that language varied between places, he postulated thatsounds changed over time, explaining the discrepancy.^[20][21]

The systematic study of Old Chinese rhymes began in the 17th century, whenGu Yanwu divided the rhyming words of theShijing into ten rhyme groups (yùnbù韻部).^[22]These groups were subsequently refined by other scholars, culminating in a standard set of 31 in the 1930s.One of these scholars,Duan Yucai, stated the important principle that characters in the same phonetic series would be in the same rhyme group,^[a] making it possible to assign almost all words to rhyme groups.^[24][12]

ModernMin dialects, particularly those of northwestFujian, show reflexes of distinctions not reflected in Middle Chinese.For example, the following dental initials have been identified in reconstructedproto-Min:^[25][26]

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 inJianyang and nearby Min varieties in northwesternFujian, where they appear as fricatives or approximants (e.g.[vlh] < *-p *-t *-k in Jianyang) or are missing entirely, while the non-softened variants appear as stops. Evidence from early loans intoMienic languages suggests that the softened stops wereprenasalized.^[27]The distinction among resonants is also reflected in someHakka varieties.^[28]

TheWaxiang dialects of westernHunan differ strongly from other Chinese varieties, preserving several distinctions and features not found in theQieyun.^[29]

Most scholars believe that Old Chinese morphemes were overwhelmingly monosyllabic, though some have recently suggested that a minority of them hadminor presyllables (but were still written with a single character).^[8][30]

Although many details of the sound system are still disputed, recent formulations are in substantial agreement on the core issues.^[31]For example, a recent synthesis gives the following inventory of initial consonants:

Uvular stops are a relatively recent proposal, replacing*ʔ,*h,*ɦ and their labialized counterparts, and are not accepted by all authors.^[33][34]Baxter and Sagart reconstruct uvular stops together with*ʔ and*ʔʷ.^[35]

Most scholars also reconstruct several initial consonant clusters.^[36][37]In most reconstructions since the 1990s, an Old Chinese syllable also has

• some representation of the distinction betweenMiddle Chinese "type-A" and "type-B" syllables, for example type A havingpharyngealized initial consonants or long vowels (see§ Type A/B distinction),^[38][39]
• an optional medial*-r- (or sometimes*-l-),^[40]
• one of sixvowels:

• an optional coda, which could be a glide*-j or*-w, a nasal*-m,*-n or*-ŋ, or a stop*-p,*-t,*-k or*-kʷ (with some authors also including*-r),
• an optional post-coda*-ʔ or*-s.

In such systems, Old Chinese, unlike later forms of Chinese, has notones. The rising and departing tones of Middle Chinese are treated asreflexes of the Old Chinese post-codas.^[41]

In the rhyme table tradition, initials were classified byplace andmanner of articulation.^[42]The initials of theQieyun are slightly different from the rhyme tables, but are generally agreed and each traditionally named with an exemplary word as follows:^[43][44]

As many potential combinations of initial and final did not occur, the designers of the rhyme tables saved space by separating finals with different patterns of co-occurrence, effectively identifying cases ofcomplementary distribution.^[53]Thus finals are split between divisions (等děng) named I, II, III and IV based on the rows of the rhyme tables in which they were placed.^[54]Most scholars believe that finals of divisions I and IV contained low back and mid front vowels respectively, while Division II has a front lax vowel. Division III is usually taken as indicating a palatal medial-j- in Middle Chinese.^[55][56]

Division III finals occur in more than half of the syllables of theQieyun.Some authors call them type B finals, with type A encompassing all the other divisions.^[57]Comparing placement in the rhyme tables with distribution in theQieyun,Li Rong subdivided division III finals into four distributional classes, which others have called pure (or independent), mixed and two kinds calledchongniu.^[58][59][60]Some of the distinctions between the resulting seven classes are found only in the rhyme tables.When considering theQieyun only, they reduce to four distinct patterns of co-occurrence with initials at various places of articulation:^[61]

The rhyme tables also classifiedQieyun finals as either 'open' (開kāi) or 'closed' (合hé), with the latter believed to indicate a medial-w- or lip rounding.^[54]The Middle Chinese-w- medial was unevenly distributed, being distinctive only after velar and laryngeal initials or before certain finals with codas-j,-n or-t.^[63]

The primary sources of evidence for the reconstruction of the Old Chinese initials are medieval rhyme dictionaries and phonetic clues in the Chinese script.

Early in the 20th century,Huang Kan proposed that the division I and IV finals, and the initials with which they occurred, namely labials, dentals, dental sibilants, velars and laryngeals, were primitive.^[64][44]

A key principle, first proposed by the Swedish sinologistBernhard Karlgren, holds that the initials of words written with the same phonetic component had a commonpoint of articulation in Old Chinese.^[65]In 1940, Karlgren published the first complete reconstruction of Old Chinese in a dictionary called theGrammata Serica, in which characters are arranged by phonetic series within rhyme groups.The 1957 revisionGrammata Serica Recensa (GSR) remains a standard reference, even though Karlgren's reconstructions have been superseded by the work of later scholars.^[66]

The uneven distribution of the Middle Chinese medial-w- is taken (followingAndré-Georges Haudricourt andSergei Yakhontov) to indicate that Old Chinese hadlabialized variants of the back initials but no labiovelar medial.^[63]The plain and labialized consonants are generally kept separate in phonetic series.^[67]The remaining occurrences of Middle Chinese-w- are believed to result frombreaking of a back vowel before these codas (see§ Vowels).^[68][69]

For example, since Middle Chinese dentals, retroflex stops and palatals, which are in complementary distribution, occur together in phonetic series, they are traced to a single Old Chinese dental series,*t-,*tʰ-,*d- and*n-, with the retroflex stops conditioned by an Old Chinese medial*-r- and the palatals conditioned by type B finals.^[70][71]The Qing philologistQian Daxin had already discovered, by studying sound glosses given byEastern Han authors, that the Middle Chinese dental and retroflex stop series were not distinguished at that time.^[72][73]

The Middle Chinese dental sibilants and retroflex sibilants are also in complementary distribution and occur together in phonetic series. They may be similarly traced to a single Old Chinese sibilant series,*ts-,*tsʰ-,*dz-,*s- and (for some authors)*z-, with the retroflex sibilants conditioned by the Old Chinese medial*-r-.^[74][75]The Middle Chinese initialsz- andzr- have restricted distributions, and many authors prefer to derive them from clusters instead of reconstructing Old Chinese*z-.^[76][77]

As Middle Chineseg- occurs only in palatal environments (type B syllables), Li attempted to derive bothg- andh- from Old Chinese*ɡ-, but had to assume irregular developments in some cases.^[78]Li Rong showed that several words with Middle Chinese initialh- were distinguished in modernMin dialects. For example,厚 'thick' and後 'after' were bothhuwX in Middle Chinese, but have velar and zero initials respectively in several Min dialects. Most authors now assume that Middle Chineseg- andh- are derived from separate Old Chinese initials, which merged in type A syllables.^[79][80][81]

The traditional view has been to project the Middle Chinese laryngeals back to Old Chinese glottal stop*ʔ- and fricatives*h- and*ɦ-, with labialized counterparts*ʔʷ-,*hʷ- and*w-.^[82]To account for the fact that Middle Chinese laryngeals are found in phonetic series with velars,Pan Wuyun proposed that these Old Chinese initials were actuallyuvular stops*q-,*qʰ- and*ɢ-, and labio-uvular stops*qʷ-,*qʷʰ- and*ɢʷ-.^[83][84]

However, there are several cases where quite different Middle Chinese initials appear together in a phonetic series.Karlgren and subsequent workers have proposed either additional Old Chinese consonants or initial consonant clusters in such cases.For example, the Middle Chinese palatal sibilants appear in two distinct kinds of series, with dentals and with velars:^[j]

• 周tsyuw (<*t-, type B) 'cycle; Zhou dynasty',彫tew (<*t-, type A) 'carve' and調dew (<*d-, type A) 'adjust'^[86][87]
• 制tsyejH (<*k-, type B) 'cut out' and猘kjejH (<*kr-, type B) 'mad dog'^[88][89]

It is believed that Old Chinese dentals in type B syllables were palatalized, unless the medial*-r- was present.Some Old Chinese velars were also palatalized in type B syllables, but the conditioning factors are only partly understood.^[90]Li proposed*Kr- as the source of palatal initials occurring in phonetic series with velars or laryngeals, found no evidence for*Pr-, and attributed thechóngniǔ distinction to the vowel.Following proposals by Pulleyblank, Baxter explainedchóngniǔ using*-r- and postulated that plain velars and laryngeals were palatalized in type B syllables with a front vowel, except when*-r-) was present.However, a significant number of palatalizations are not explained by this rule.^[91][92][93]

Similarly, it is proposed that the*-r- medial could occur after labials and velars, complementing the instances proposed as sources of Middle Chinese retroflex dentals and sibilants, to account for such connections as:^[94][95][k]

• 筆pit (<*pr-, type A) 'writing pencil' and律lwit (<*br-, type A) 'law; rule'^[97][98]
• 監kaem (<*kr-, type A) 'look at' and藍lam (<*ɡr-, type A) 'indigo'^[99][100]

Thus the Middle Chinese laterall- is believed to reflect Old Chinese*-r-.Old Chinese voiced and voiceless laterals*l- and*l̥- are proposed to account for a different group of series, mixing palatals withd- andth- (but nott-), such as

• 脫dwat (<*l-, type A) andthwat (<*l̥-, type A) 'peel off',悅ywet (<*l-, type B) 'pleased' and說sywet (< OC*l̥-, type B) 'speak'^{[101][102][l]}

The divergentWaxiang dialect of westernHunan has a lateral initial in many of these words.^[103]This treatment of the Old Chinese liquids is further supported by Tibeto-Burman cognates and by transcription evidence.For example, the name "Alexandria" (referring to one ofmany cities so named) was transcribed in theBook of Han chapter 96A as ⟨烏弋山離⟩, which is reconstructed as*ʔa-ljək-srjan-rjaj.^{[104][105][106][107]}

Voiceless nasal initials*m̥-,*n̥- and*ŋ̊- are proposed (followingDong Tonghe andEdwin Pulleyblank) in series such as:^[108]

• 墨mok 'ink' and黑xok (<*m̥-) 'black'^[109][110]
• 難nan 'difficult' and灘than (<*n̥-) 'foreshore'^[111][112]
• 虐ngjak 'cruel' and謔xjak (<*ŋ̊-) 'to ridicule'^[113][114]

Clusters*sn- and so on are proposed (following Karlgren) for alternations of Middle Chinese nasals ands- such as

• 如nyo (<*n-, type B) 'resemble' and絮sjoH (<*sn-, type B) 'raw silk'^[115][116]

Other cluster initials, including*s with stops or stops with*l, have been suggested but their existence and nature remains an open question.^[117][118]

Proto-Min *nh and *lh are distinct from the Old Chinese initials*n̥ and*l̥, which both yield Middle Chineseth (orsy in palatal environments).^[119][120]The distinctions found in Proto-Min initials are assumed by most workers to date from the Old Chinese period, but they are not reflected in the widely accepted inventory of Old Chinese initials given above.The Old Chinese antecedents of these distinctions are not yet agreed, with researchers proposing a variety of consonant clusters.^[121]Baxter and Sagart derive the additional aspirated initials from consonant clusters and the softened stops fromminor syllables.^[122]

Although all authors agree that the difference between type A and B syllables reflects a real Old Chinese phonological distinction of some sort, there is no agreement on its phonetic form.^{[123][124][125]}The traditional approach was to project the Middle Chinese palatal medial-j- back to a medial*-j- in Old Chinese, but many recent authors argue that the absence or presence of the Middle Chinese medial reflects something different in Old Chinese.Evidence includes the use of type B syllables to transcribe foreign words lacking any such medial, the lack of the medial in Tibeto-Burman cognates and modern Min reflexes, and the fact that it is ignored in phonetic series.^[126][127]The distinction has been variously ascribed to:

• the presence or absence of a prefix. Jakhontov held that type B reflected a prefix*d-,^[124][57] while Ferlus suggested that type A arose from an unstressed prefix*Cə- (aminor syllable), which conditioned syllabictenseness contrasting with laxness in type B syllables.^[128]
• a length distinction of the main vowel. Pulleyblank initially proposed that type B syllables had longer vowels.^[129] Later, citing cognates in other Sino-Tibetan languages, Starostin and Zhengzhang independently proposed long vowels for type A and short vowels for type B.^{[130][131][132]} The latter proposal might explain the description in some Eastern Han commentaries of type A and B syllables ashuǎnqì緩氣 'slow breath' andjíqì急氣 'fast breath' respectively.^[34]
• aprosodic stress-based distinction, as later proposed by Pulleyblank,^[129] in which type B syllables were stressed in the first mora, while type A syllables were stressed on the second^[133]
• pharyngealization of the initial consonant.Norman suggested that type B syllables (his class C), which comprised over half of the syllables of theQieyun, were in factunmarked in Old Chinese. Instead, he proposed that the remaining syllables were marked by retroflexion (the*-r- medial) or pharyngealization, either of which prevented palatalization in Middle Chinese.^[134] Baxter andSagart have adopted a variant of this proposal, reconstructing pharyngealized initials in all type A syllables.^[135]

It is possible to account for the combinations of initials and finals of theQieyun by combining these initials with and Old Chinese medial *-r- and the Old Chinese counterpart of the type A/B distinction as in the following table.^[91][136]

Here*P,*T,*TS,*K and*Q stand for consonant classes in Old Chinese.Some authors also add a medial*-l-.^[137][40]

Assuming that rhyming syllables had the same main vowel, Li Fang-Kuei proposed a system of four vowels*i,*u,*ə and*a.He also included three diphthongs*iə,*ia and*ua to account for syllables that were placed in rhyme groups reconstructed with*ə or*a but were distinguished in Middle Chinese.^[138]In the late 1980s,Zhengzhang Shangfang,Sergei Starostin and William Baxter (followingNicholas Bodman) independently argued that these rhyme groups should be split, refining the 31 traditional rhyme groups into more than 50 groups corresponding to a six-vowel system.^{[139][140][141][142][143]}Baxter supported this thesis with a statistical analysis of the rhymes of theShijing, though there were too few rhymes with codas*-p,*-m and*-kʷ to produce statistically significant results.^[144]

The following table illustrates these analyses, listing the names of the 31 traditional rhyme groups with their Middle Chinese reflexes and their postulated Old Chinese vowels in the systems of Li and Baxter. Baxter and Sagart (2014)'s system is virtually unchanged vowel-wise from Baxter (1992), but they also state that Old Chinese had*-ik and*-iŋ rhymes that merged with other rhyme classes in Middle Chinese.Following the traditional analysis, the rhyme groups are organized into three parallel sets, depending on the corresponding type of coda in Middle Chinese.For simplicity, only Middle Chinese finals of divisions I and IV are listed, as the complex vocalism of divisions II and III is believed to reflect the influence of Old Chinese medials*-r- and*-j- (see previous section).^[145][m]

There has been much controversy over the relationship between final consonants and tones, and indeed whether Old Chinese lacked the tones characteristic of later periods, as first suggested by theMing dynasty scholarChen Di.^[o]

Thefour tones of Middle Chinese were first described byShen Yue around AD 500.They were the 'level' (平píng), 'rising' (上shǎng), 'departing' (去qù), and 'entering' (入rù) tones, with the last category consisting of the syllables ending in stops (-p,-t or-k).^[148]Although rhymes in theShijing usually respect these tone categories, there are many cases of characters that are now pronounced with different tones rhyming together in the songs, mostly between the departing and entering tones.This ledDuan Yucai to suggest that Old Chinese lacked the departing tone.Wang Niansun (1744–1832) andJiang Yougao (d.1851) decided that the language had the same tones as Middle Chinese, but some words had later shifted between tones, a view that is still widely held among linguists in China.^[149][150]

Karlgren also noted many cases where words in the departing and entering tones shared a phonetic element within their respective characters, e.g.

• 賴lajH 'depend on' and剌lat 'wicked'^[151]
• 欬khojH 'cough' and刻khok 'cut; engrave'^[152]

He suggested that the departing tone words in such pairs had ended with a final voiced stop (*-d or*-ɡ) in Old Chinese.^[153]Being unwilling to split rhyme groups, Dong Tonghe andLi Fang-Kuei extended these final voiced stops to whole rhyme groups.The only exceptions were the歌 and祭 groups (Li's*-ar and*-ad), in which the traditional analysis already distinguished the syllables with entering tone contacts.The resulting scarcity of open syllables has been criticized ontypological grounds.^[154]Wang Li preferred to reallocate words with connections to the entering tone to the corresponding entering tone group, proposing that the final stop was lost after a long vowel.^[155]

Another perspective is provided byHaudricourt's demonstration that the tones ofVietnamese, which have a very similar structure to those of Middle Chinese, were derived from earlier final consonants.The Vietnamese counterparts of the rising and departing tones derived from a final glottal stop and*-s respectively, the latter developing to a glottal fricative*-h.These glottal post-codas respectively conditioned rising and falling pitch contours, which became distinctive when the post-codas were lost.^[156]Haudricourt also suggested that the Chinese departing tone reflected an Old Chinese derivational suffix*-s.The connection with stop finals would then be explained as syllables ending with*-ts or*-ks, with the stops later disappearing, allowing rhymes with open syllables.^[157]The absence of a corresponding labial final could be attributed to earlyassimilation of*-ps to*-ts.Pulleyblank supported the theory with several examples of syllables in the departing tone being used to transcribe foreign words ending in-s into Chinese.^{[158][159][160]}

Pulleyblank took Haudricourt's suggestion to its logical conclusion, proposing that the Chinese rising tone had also arisen from a final glottal stop.^[161]Mei Tsu-lin 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, e.g.Wenzhounese and some Min dialects.^[162]In addition, most of the entering tone words that rhyme with rising tone words in theShijing end in-k.^[163]

Together, these hypotheses lead to the following set of Old Chinese syllable codas:^[164]

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.^[165]

To account for phonetic series and rhymes in which MC-j alternates with-n,Sergei Starostin proposed that MC-n in such cases derived from Old Chinese*-r.^[166]Other scholars have suggested that such contacts are due to dialectal mixture, citing evidence that*-n had disappeared from eastern dialects by theEastern Han period.^[167]Baxter and Sagart cite transcription evidence for*-r, arguing that it became*-n in Western Old Chinese dialects and*-j in Eastern ones.^[168]

• Historical Chinese phonology

Wikibooks has a book on the topic of:Character List for Karlgren's GSR

Tutorials

• "Introduction to Chinese Historical Phonology",Guillaume Jacques, European Association of Chinese Linguistics Spring School in Chinese Linguistics,International Institute for Asian Studies, Leiden, March 2006.
• "Summer School on Old Chinese Phonology"^{[permanent dead link]} (videos), William Baxter and Laurent Sagart,École des Hautes Études en Sciences Sociales, Paris, July 2007.

Databases of reconstructions

• StarLing database, by Georgiy Starostin.
• The Baxter-Sagart reconstruction of Old Chinese.
• 上古音系- 韻典網
• 小學堂上古音 - 中央研究院
• 李方桂上古音韻表

• Old Chinese
• Sino-Tibetan phonologies

• Articles containing Chinese-language text
• Pages with plain IPA
• Articles containing traditional Chinese-language text
• Articles with short description
• Short description is different from Wikidata
• Use dmy dates from October 2020
• CS1 uses Chinese-language script (zh)
• CS1 Chinese-language sources (zh)
• CS1 Russian-language sources (ru)
• CS1: abbreviated year range
• All articles with dead external links
• Articles with dead external links from July 2025
• Articles with permanently dead external links