Late Middle Chinese phonology as recorded in the Yunjing (韻鏡) — an introduction

Table 1 of the Yunjing

Last updated: 2024–12–11

The Yunjing (韻鏡) is a book of rime tables that grouped together Chinese characters that rhymed. It was published as early as 1161 (Southern Song), though exactly when it was compiled, and who compiled it, is unknown.

Unfortunately, exact phonetic values are not explicitly given, and some of them remain open to speculation. But by using pronunciations in modern Sinitic and non-Sinitic languages, as well as other pieces of evidence, it is possible to make educated inferences about these phonetic values.

In this article, I will describe the layout of the Yunjing, discuss its terminology, and provide a reconstruction of phonetic values in places where they are not explicitly given. Some of these phonetic values are my own proposals, but the arguments supporting them are too long to include here. I will lay out the case for them in later articles. You can find a list of the topics I plan to discuss at the bottom of this article; these topics will link to any completed articles.

A short history

The earliest formal record of the pronunciations of Chinese characters, and the point commonly regarded as the start of Middle Chinese, was the Qieyun (切韻 ‘cut rhymes’), compiled in 601 AD (Sui Dynasty) by Lu Fayan (陸法言). The Qieyun grouped together homophones and rhyming words, and described the pronunciation of each word (which were all monosyllabic or sesequisyllabic) using a method called fanqie 反切: two other words were given, one with the same syllabic onset (or initial) as the pronunciation being described, and another with the same rime (or final) and tone. For example, the fanqie of 東 [*tʊŋ˥] ‘east’ was 德紅, because 德 [*tək̚] ‘virtue’ had the same initial, and 紅 [*ɦʊŋ˥] had the same final and tone. The rhyme groups were grouped by tone.

Later came the Guangyun (廣韻 ‘extensive rhymes’), a revised and expanded version of the Qieyun completed in 1008 (Northern Song). The Guangyun was then followed by the Jiyun (集韻 ‘collected rhymes’) in 1037 (Northern Song), a revised and expanded version of the Guangyun.

Rime tables likely developed alongside rime dictionaries, and were more linguistically-sophisticated. The Yunjing is one of the earliest known set of rime tables, along with the Qiyin Lüe (七音略 ‘summary of the seven sounds’), the earliest known version of which also dates to 1161 AD. Both sets of rime tables are believed to be versions of a common source from before the Song Dynasty.

Unlike the rime dictionaries, which all relied on fanqie to describe pronunciations, the rime tables gives explicit descriptions of the phonology of initials. It also groups together certain previously non-rhyming finals of the same tone into sets of four rows.

The layout

The Yunjing contains 43 rime tables, each with 23 columns and 16 rows. The first table is shown at the top of the article as an example.

All the characters in each table share the same main vowel, and the same place of articulation (bilabial, alveolar, or velar) in their codas, if any. The larger characters in the headings on the left are the head words of rhyme groups (groups of words that rhyme) from earlier rime dictionaries like the Guangyun. Each character in the table is generally also the word that heads a homophone group (a group of words with the exact same pronunciation) in these rime dictionaries, though occasionally a different word is selected. Pronunciations with no known characters are marked with a circle.

Following the traditional analysis of Chinese syllabic structure, syllables are broken down into an initial and a final. The initial is the initial consonant, and corresponds to the onset in modern phonology. The final refers to everything after the initial, and can be further broken down into the medials, the main (or syllabic) vowel, and the coda. In modern phonology, it corresponds to the rime.

The medial is a sequence of one or more approximants before the main vowel. This differs somewhat from modern phonological analyses, where a medial approximant would usually be analyzed as either the non-syllabic part of a rising diphthong, or the last part of a complex consonant cluster.

For example, the Yunjing pronunciation of 亮 [*lɨ̯jʌŋ˥˩] ‘bright’ would be analyzed as

initial: l
final: ɨ̯jʌŋ
tone: ˥˩

final
medials: ɨ̯j
main vowel: ʌ
coda: ŋ

Initials

The 23 columns are for the initial consonants (also known as the onset). They are divided into six groups:

1. labials (唇音 “lip sounds”)
2. coronals (舌音 “tongue sounds”)
3. velars (牙音 “back teeth sounds”)
4. sibilants (齒音 “front teeth sounds”)
5. laryngeals (喉音 “throat sounds”)
6. liquids (舌齒音 “tongue/front teeth sounds”)

Each column is also marked for voicing and aspiration:

1. plain/tenuis (unvoiced, unaspirated) (清 “clear”)
2. aspirated (次清 “less clear”)
3. voiced (濁 “muddy”)
4. sonorant (清濁 “clear-muddy”)

Columns in the bilabial, velar, laryngeal, and liquid categories represent a single initial each, while columns in the coronal and sibilant categories represent alveolar consonants on Rows 1 and 4, while on Rows 2 and 3 coronals are retroflex and sibilants are post-alveolar.

A chart at the beginning of the Yunjing introduces the 36 initials, and describes how they fit into the 23 columns. Each initial is represented by a character with that initial:

The initials chart in the Yunjing

Below is a chart of the phonetic values of the 36 initials that I propose, along with the characters used to represent them in the Yunjing:

The 36 initials in the Yunjing, each represented by a character with the initial and its phonetic value

Most of these initials are similar or identical to those proposed by previous linguists such as Pulleyblank and Baxter, but I have also proposed a few revisions, specifically in the ‘light-lip’ labials [*ɸʷ, *ɸʷː, *βʷ, *w], the ‘true front-tooth’ sibilants [*tʃ, *tʃʰ, *dʒ, *ʃ, *ʒ], the ‘half-sibilant’ sonorant [ɹ̝], and the sonorant laryngeal [*ʔ̞]. I will justify these proposals in later articles.

Medials

There are four medials in the Yunjing: [*-w-], [*-j-], [*-ɥ-], and [*-ɨ̯-]. The [*-ɥ-] medial results from combining [*-w-] with [*-j-].

Of the four medials, modern Mandarin has retained three, [*-w-], [*-ɥ-], and [*-j-]. The fourth, [*-ɨ̯-], is my proposal.

The phonetic values of the four rows have been the subject of some debate. I propose that for bilabials, velars, laryngeals, and liquids, the four rows indicate the presence or absence of two medials, [*-j-] and [*-ɨ̯-]; for coronals and sibilants, [*-ɨ̯-] is replaced by tongue-retraction in the initial. The [*-ɨ̯-] medial, retroflex coronals, and post-alveolar sibilants developed from an older [*-ɹ̈-] medial that was already obsolete by the time of the Yunjing.

This leads to the following interpretation of the four rows:

Row 1: No medial
Row 2: Previously [*-ɹ̈-]. Preceding coronals became retroflex; preceding sibilants became post-alveolar; [*-ɹ̈-] became [*-ɨ̯-] elsewhere
Row 3: Previously [*-ɹ̈j-]. Preceding coronals became retroflex, and preceding sibilants became post-alveolar, both followed by [*-j-]; [*-ɨ̯j-] elsewhere
Row 4: [*-j-] medial

The presence or absence of the [*-w-] medial is indicated for each table: the label 開 ‘open’ indicates its absence, 合 ‘closed’ indicates its presence, and 開合 ‘open-closed’ indicates that it is optional.

Main vowels and codas

Each table is associated with either one or two finals. Finals with either [*-w], [*-j], or no coda apply to the entire table. Finals with the nasal codas [*-m], [*-n], or [*-ŋ] share a table with another final that shares the same main vowel, but has an unreleased stop coda with the same place of articulation, [*-p̚], [*-t̚], or [*-k̚]. Unless stated otherwise, the bottom four rows of the table are reserved for words with stop codas; the exceptions are Tables 9–10 and Tables 13–14, where they are used for another departing tone rhyme group from the Guangyun (indicated by the phrase 去聲寄此 ‘departing tone goes here’).

This practice of grouping finals with nasal codas together with finals with homorganic unreleased stop codas stems from the Chinese linguistic tradition of treating unreleased stop coda as a tone, known as the 入聲 ‘entering tone.’ While this may seem unusual to anyone not familiar with Chinese, the traditional analysis makes some sense in view of the historical relationship between Chinese tones and codas: tones in Chinese words without stop codas are believed to have originally developed from codas.

The phonetic values of the finals are not indicated directly in the Yunjing. We can, however, use the modern forms of words in both Sinitic and non-Sinitic languages, as well as rhyming words in Tang poetry, to infer the likeliest sounds that the finals could have contained.

Additionally, each table is labeled either 内 ‘inner’ or 外 ‘outer.’ The distinction between inner and outer tables has yet to be correlated with any other single property, and it is made differently in different rime books, so its exact meaning remains elusive. However, it very roughly corresponds to a contrast between high and low vowels. In the Yunjing, all the main vowels in Row 2 words with a non-sibilant initial and a rhotic medial are low ([*-æ] and [*-ɔ]); all of the tables containing these words, except for Table 29 ([*-æ]), are marked ‘outer’ in the Yunjing. This still leaves several other outer tables unaccounted for, however, namely Tables 17–20 ([*-(w)ɤn]), which have a high vowel. I will leave this topic for a potential future article.

Taken as a whole, the evidence suggests that there were six main vowels in the Yunjing, [*i, *æ, *ʌ, *ɔ, *ɤ, *o] (vowels before colons are unrounded; those after colons are rounded):

The six main vowels in the Yunjing

Of these six vowels, [*ɔ] is the r-coloured allophone of [*o], while [*æ] is the r-coloured allophone of [*ʌ] and [*ɤ]. This reduces the vowels to four phonemes, /*i, *ʌ, *o, *ɤ/.

In sum, I propose these reconstructed finals for the 43 tables of the Yunjing, along with their ‘outer’ or ‘inner’ label:

Tables 1–2: [*-oŋ] (inner)
Table 3: [*-ɔŋ] (outer)
Tables 4–10: [*-i] (inner)
Tables 11–12: [*-o] (inner)
Tables 13–16: [*-æj] (outer)
Tables 17–20: [*-ɤn] (outer)
Tables 21–24: [*-æn] (outer)
Tables 25–26: [*-æw] (outer)
Tables 27–28: [*-ʌ] (inner)
Tables 29–30: [*-æ] (29: inner; 30: outer)
Tables 31–32: [*-ʌŋ] (inner)
Tables 33–36: [*-æŋ] (outer)
Table 37: [*-ɤw] (inner)
Table 38: [*-ɤm] (inner)
Table 39–41: [*-æm] (outer)
Tables 42–43: [*-ɤŋ] (inner)

Again, some of these values differ from those in common reconstructions. I will discuss these in later articles.

Tones

In the rime tables, the 16 rows are divided into four sections, each with four rows. The first three sections correspond to the three lexical tones that distinguish syllables not ending in a stop coda ([p̚, t̚, k̚]). The top section is for the level (平) tone (˥), the second section for the rising (上) tone (˩˥), and the third section for the departing (去) tone (˥˩).

The exact pitches of the three tones are speculative. In any case, since pitches did not provide contrast at the time of the Yunjing, they are of secondary importance relative to the tonal contours.

As previously mentioned, the fourth section (except in Tables 9–10 and 13–14) is for the so-called entering (入) tone, which is not strictly a tone, but contains syllables ending in a stop coda.

Conveniently, each of these four Chinese terms has the tone it named: 平 [*bɨ̯jæŋ˥], 上 [ʒjʌŋ˩˥], 去 [kʰɨ̯jo˥˩], and 入 [ɹ̝ɨ̯jɤp̚]. Additionally, 平 and 上 appear to also describe the contour of the tones they name (the relationships between tonal contours and the meanings of 去 and 入 are more obscure).

Specific topics

Below is a list of topics that I plan on discussing in more detail, including justifications for any original proposals I have made.

This list is subject to change. Topics in the list will become links to any completed articles.

1. The four rows of the Yunjing
2. The lenition of bilabial initials
3. Tongue retraction in sibilant, coronal, and rhotic initials
4. The creaky-voice initial
5. Finals in the Yunjing
6. Tones in the Yunjing
7. Inner and outer tables in the Yunjing