GX202411 Tangut Transcription System

GX202411 Tangut Transcription System

This document provides an overview of the GX202411 transcription system for the Tangut language. It is a work in progress (see Changelog), with the most recent major updates made on:

2025-04-07

To cite the GX202411 system itself (rather than this overview), please refer to Gong (2025):

  • Gong, Xun (2025). “Tone values of Tangut”. Bulletin of Chinese Linguistics 18.1.

If you find any errors or have suggestions for improvement, please contact:

Purpose of this Document

GX202411 is one of a series of transcription systems associated, so to speak, with the ‘Xun Gong reconstruction’, incorporating the author’s key hypotheses regarding Tangut phonology, particularly in relation to uvularization (Gong 2020) and nasal preinitials (Gong 2021). A primary objective of this system is to provide a more precise set of conventions while improving ease of use, not least for scholars whose primary focus lies outside Tangut historical linguistics.

This document serves as a practical guide to the GX202411 transcription system. It explains the symbols used, the underlying phonological framework, and provides pedagogical conventions for pronunciation. Although it discusses the system’s relationship to native Tangut dictionaries and compares it with other reconstruction/transcription systems, this document does not aim to justify the specific choices. For detailed arguments regarding the core hypotheses (uvularization and nasal preinitials), readers are referred to the author’s publications. Some specific points presented here are results of ongoing research; detailed justifications will be provided in future publications.

It is important to note that the IPA values enclosed in square brackets [], while rich in phonetic detail, are not meant to represent the author’s view on the reconstructed phonetic value. Instead, they serve as indicative guides, often reflecting the author’s personal pronunciation habits and representing a plausible intersection of reconstructed phonetic possibilities and pedagogical practicality.

Transcription Symbols

The following tables of transcription symbols also function as a table of contents for the subsequent sections.

Initial consonants

Labial Dental Alveolar Palatal Retroflex Velar Uvular Glottal
p [p] t [t] ts [ts] tś [tɕ] tṣ [tʂ] k [k] q [q] Ø [ʔ]
𗟱 pə¹ 𗓱 te¹ 𗥤 tse² 𗵘 tśa¹ 𗉋 tṣao̱ṃ¹ 𗑱 ka² 𗟔 qo̱¹ 𗈪 a¹
ph [pʰ] th [tʰ] tsh [tsʰ] tśh [tɕʰ] tṣh [tʂʰ] kh [kʰ] qh [qʰ]
𗉣 phi¹ 𗋕 tha¹ 𗱱 tsha² 𗉘 tśhə¹ 𗽇 tṣhaa̱¹ 𗯴 khu¹ 𘂤 qha̱¹
b [b] d [d] dz [dz] dź [dʑ] dẓ [dʐ] g [ɡ] ġ [ɢ]
𘞆 bi¹ 𘟣 du¹ 𗡅 dzi¹ 𗱠 dźe¹ 𘓟 dẓae̱² 𘒏 ga¹ 𗦬 ġu̱²
f [f] lh [ɬ] s [s] ś [ɕ] ṣ [ʂ] h [h] h [h]
𘋇 fi¹ 𗽀 lhə² 𗜍 sa¹ 𗀓 śe² 𗀚 ṣao̱¹ 𘜤 ha¹ 𗥑 hu̱¹
ll [ɮ] z [z] ź [ʑ] ẓ [ʐ] gh [ɣ] ġh [ʁ]
𘜶 lleh² 𘄄 zi¹ 𗋾 źu² 𗖣 rẓai̱r¹ 𗐜 gha² 𗼜 ġhai̱²
v [ʋ] l [l] y [j] r [ɻ] w [ɦw]
𘃡 vi¹ 𘟀 le² 𘝵 ye¹ 𘆝 rer¹ 𘃽 wo̱²
m [m] n [n] ṇ [ɳʵ] ŋ [ɴ]
𗴺 ma¹ 𘟙 ne² 𘊄 ṇoṃ² 𗧓 ŋa̱²

Vowels and Uvularization

The system postulates six vowel phonemes (/a/, /e/, /i/, /o/, /u/, /ə/), each exhibiting a contrast between a plain (Grade III) and a uvularized (Grade I, /ʶ/) variant. Uvularized vowels are marked with a subscript line (a̱, e̱... U+0331). Grade II syllables are analysed as containing a Grade II medial (-a-), which occurs only with phonologically uvularized vowels.

/a/ /e/ /i/ /o/ /uo/ /u/ /ə/
/V/ (Grade III) a [ɐ] e [e] i [i] o [ɵ] uo [ʉ͡ɵ] u [ʉ] ə [ɨ]
𗴺 ma¹ 𗓱 te¹ 𘃡 vi¹ 𗤾 lo² 𘔼 nuo¹ 𗨁 phu² 𘉑 mə¹
/Vʶ/ (Grade I/II) a̱ [ɑʶ] e̱ [ɛ̠ʶ] i̱ [ɪʶ] o̱ [ɔʶ] --- u̱ [ʊʶ] ə̱ [ʌʶ]
𗓰 na̱¹ 𗑉 me̱¹ 𗾔 bi̱² 𘊳 bo̱¹ 𘕰 phu̱² 𗹦 mə̱¹

Meta-notation and Tones

  • ?: Denotes unknown tone or uncertain overall pronunciation; by pedagogical convention, pronounced as Tone 2 - 𘝦 dźə?
  • ¹: Tone 1 (𗗔 nye¹ = 平 píng) high-falling tone HL in monosyllables (Gong 2025)
  • ²: Tone 2 (𗨁 phu² = 上 shàng) mid tone M in monosyllables (Gong 2025)

Segmental features

Pan-syllabic features

Syllable examples

The following examples illustrate how to identify various segmental and suprasegmental features:

Example Pre-initial Initial Medial(s) Vowel Coda Suprasegmental Features
𗅢 nsi̱w² n- s i̱ w Tone 2 (²), Uvularized (i̱)
𗏆 hwaa̱ṃ¹ h -w- -a- a̱ ṃ Tone 1 (¹), Uvularized (a̱)
𗫨 rdwi̱wr² d -w- i̱ w Tone 2 (²), Uvularized (i̱), Rhotacized (r- -r)

Tones

In line with the Chinese phonological tradition, native Tangut scholars recognized two primary tonal categories: 𗗔 nye¹ ‘flat’ (平 píng) and 𗨁 phu² ‘high/upper’ (上 shàng). In contemporary transcription systems, including this one, these are commonly referred to as Tone 0 and Tone 2, indicated respectively by the superscripts ¹ and ².

For characters with an uncertain tonal category or unknown overall pronunciation, a widely accepted convention across transcription systems is to use a meta-notational symbol in place of the tone mark, practically treating unknown as another tonal category. This system follows this practice and employs ?. A well-known example is 𘝦 dźə? meaning ‘action, deed’, which is part of the frequently used ergative marker 𘝦𘃡 dźə?vi¹. As a pedagogical guideline, this system recommends pronouncing such characters with Tone 2.

Monosyllabic Tone Values

Whether or not the student of Tangut is familiar with tonal languages, to avoid confusion, it is recommended to make a distinct effort to clearly differentiate the monosyllabic tonal values. The commonly accepted view on tonal values (Nishida 1964; Arakawa 1999) posits ‘1 = High, 2 = Low’. However, Gong (2025) strongly argues for a revised Tone 1 value as high-falling. Therefore, the recommended pronunciation for monosyllables in this system is as follows:

  • Tone 1 (𗗔 nye¹ = píng): high-falling contour HL [˥˨]
  • Tone 2 (𗨁 phu² = shàng): mid-level contour M is considered by Gong (2025) to be the most likely, but other possibilities are also plausible. In practice, any pitch shape that is neither high nor falling is acceptable.
  • Tone 1: For practical purposes, Tone 1 can be approximated by Standard Mandarin Tone 4 (e.g. 骂 mà). However, one should strive to avoid the creaky voice often heard at the end of Mandarin T4 to prevent confusion with Tense Vowels -h.
  • Tone 2: A range of realizations can be considered suitable; choose the most comfortable pronunciation that is neither high nor falling. For Standard Mandarin speakers, any tone other than T4 can serve as an acceptable approximation: Tone 1 (妈 mā, preferably slightly lower, with a Northeastern, Húnán, or Malaysian accent); Tone 2 (麻 má); or Tone 3 (马 mǎ, avoiding creaky voice if possible). The author’s personal preference is a mid-level tone with a slight rise near the end [˧˧˦], a pitch shape attested in many Rgyalrongic languages.
Tone 1 (HL) Tone 2 (M)
𘑶 tse¹ ‘other’ 𗥤 tse² ‘to understand’
𗁮 tśhi¹ ‘flesh’ 𗺉 tśhi² ‘classics’
𗅋 mi¹ neg. prefix 𗼇 mi² ‘Tangut’
𘐏 tih¹ ‘to put’ 𘏚 tih² ‘place’

Pitch Shape in Polysyllables

The realization of tones in polysyllabic Tangut words is uncertain and may forever remain unknown. In practice, however, simply concatenating monosyllabic tone values would likely be incorrect and result in highly unnatural speech flow. Based on the phonological typology of related languages, this system proposes the following pedagogical convention:

  • Pitch peak location: Within a polysyllabic prosodic unit, excluding enclitics:
    • The leftmost Tone 1 syllable is considered prominent, thus carrying the highest pitch of the word; syllables preceding and following it should be lower.
    • If all syllables are Tone 2, the rightmost Tone 2 syllable is considered prominent and should carry the relative pitch peak.
  • Contour preservation on rightmost prominent syllable: If the prominent syllable (determined as above) is the rightmost non-enclitic syllable of the word, it should retain its full monosyllabic tonal contour (HL for Tone 1, M for Tone 2).
  • Enclitics: Certain particles and grammatical suffixes are treated as unstressed enclitics and should be pronounced either with a low pitch or by propagating the pitch contour of the preceding non-enclitic syllable. Verb person suffixes, such as 𗧓 -ŋa̱² and 𘂆 -tsə¹, are always treated as enclitics; no stipulation is made for other particles and suffixes within this pedagogical convention.

These rules allow considerable flexibility regarding pitch lowering and contour compression on non-prominent syllables. As long as the pitch peak correctly falls on the leftmost Tone 1 syllable, the overall pitch shape can, and should, be adapted to the speaker’s native language habits and personal preferences. For instance, the author’s personal habit is such that the monosyllabic contour on the final syllable is retained, albeit somewhat reduced, even when that syllable is not prominent (i.e. preceded by a Tone 1); thus, a Tone 1–1 sequence like 𗯨𗖻 rur¹puh¹ might be pronounced H.ML. However, pronunciations with the second syllable fully lowered, such as HL.L or H.L, are also perfectly conformant with this pedagogical convention.

Longer nominal compounds need not be pronounced as a single prosodic unit. They can be divided into multiple units based on semantic grouping, with tone patterns determined independently for each unit. For example, the Tangut equivalent for tathāgata-arhat-samyaksaṃbuddha, 𗌮𗆐𗖵𗟭𗣼𗳦𗫨 rmor¹lleh²bu¹vwo̱²tśha²qa̱¹rdwi̱wr², can be segmented for easier pronunciation, e.g. 𗌮𗆐𗖵𗟭𗣼𗳦𗫨 rmor¹lleh² bu¹vwo̱² tśha²qa̱¹rdwi̱wr²..

Examples:

Pattern Recomm. Pitch Example Notes
1–1 H.x 𗯨𗖻 rur¹puh¹ Initial Tone 1
1–2 H.x 𗟱𗾞 pə¹nnə² Initial Tone 1
2–1 x.HL 𗦎𘈧 ra̱r²nde̱¹ Leftmost Tone 1 is final; contour retained
2–2 x.M 𘝶𗦜 the²so² All Tone 2; final prominent
2–1–1 x.H.x 𘜶𗈁𗤶 lleh²nvu¹nne¹ Leftmost Tone 1 is medial; highest pitch
2–2–1 x.x.HL 𘋢𗥤𗵘 dẓaa̱ṃ²tse²tśa¹ Leftmost Tone 1 is final
1–2=σ H.x(=x) 𗈪𗶹𗧓 a¹śi²ŋa̱² Enclitic
2–1=σ x.HL(=x) 𘙇𗶷𘂆 rər²śə¹tsə¹ Enclitic

Vowels and Contrastive Uvularization

Gong (2020) reconstructs for the Tangut language a basic six-vowel system: /a/, /e/, /i/, /o/, /u/, and /ə/. Each of these vowels has two variants: a plain form (Grade III) and a uvularized form (Grade I, phonologically /ʶ/). Uvularized vowels are denoted by a subscript line (a̱, e̱, etc. U+0331). While Grade II syllables are also considered phonologically uvularized, they are differentiated from Grade I syllables by the presence of the Grade II medial (-a-), which will be discussed later. Additionally, the system recognizes one unanalysable diphthong, uo.

Uvularization in Tangut, represented as /ʶ/, is a secondary articulation applied to vowels. It involves pronouncing a slight uvular approximant or fricative [ʁ], similar to the ‘r’ sound in French or Standard German, simultaneously throughout the vowel's duration. Evidence suggests that Tangut uvularization, like similar phenomena in other languages, is linked to Retracted Tongue Root (RTR) and backness, which significantly affect vowel quality. The IPA values in brackets indicate one possible phonetic realization. Importantly, these values represent the auditory perception resulting from applying uvularization to the plain counterpart (e.g. pronouncing [e] with uvularization added subsequently naturally sounds like [ɛ̠ʶ]), rather than indicating a target articulation of [ɛ̠ʶ] to which uvularization is then added.

To grasp the contrast intuitively, practise the vowel pairs with velar (k-) vs uvular (q-) initials. Note that syllables that would have had the shape *qi̱ have undergone epenthesis of the Grade II medial -a- in the phonological history of Tangut, becoming *qai̱, thus rendering the i/i̱ pair inapplicable for this exercise.

Plain (Grade III) Example Uvularized (Grade I) Example
a [ɐ] 𗑱 ka² [kɐ˧˧˦] a̱ [ɑʶ] 𗈜 qa̱² [qɑʶ˧˧˦]
e [e] 𗿈 khe² [kʰe˧˧˦] e̱ [ɛ̠ʶ] 𗉫 ŋqhe̱¹ [ɴqʰɛ̠ʶ˥˨]
i [i] 𘈷 gi² [ɡi˧˧˦] i̱ [ɪʶ] (𗾔 bi̱² [bɪʶ˧˧˦])
o [ɵ] 𘓯 khoṃ¹ [kʰɵw̃˥˨] o̱ [ɔʶ] 𗟔 qo̱¹ [qɔʶ˥˨]
u [ʉ] 𗯴 khu¹ [kʰʉ˥˨] u̱ [ʊʶ] 𗼒 qhu̱¹ [qʰʊʶ˥˨]
ə [ɨ] 𘙌 kə¹ [kɨ˥˨] ə̱ [ʌʶ] 𗛖 qə̱¹ [qʌʶ˥˨]

Vowels a [ɐ] and a̱ [ɑʶ]

  • Plain a [ɐ]: Any raised and/or fronted a-like vowel. The actual quality likely exhibited a strong degree of variation within the range [ɐ ~ ɛ ~ æ ~ a].
  • Uvularized a̱ [ɑʶ]: Uvularized low central/back unrounded vowel.
Plain a Uvularized a̱
𗧤 pha¹ [pʰɐ˥˨] ‘to cut’ 𗂙 pha̱¹ [pʰɑʶ˥˨] ‘other’
𗐜 gha² [ɣɐ˧˧˦] ‘to cover’ 𘕿 ġha̱² [ʁɑʶ˧˧˦] locative part.
𗋕 tha¹ [tʰɐ˥˨] ‘that’ 𗢳 tha̱¹ [tʰɑʶ˥˨] ‘Buddha’

The canonical reference for vowel quality is the contrast in the vowel ā between ك kāf and ق qāf in most varieties of Arabic.

Those whose native language distinguishes two a-like sounds can typically apply this distinction to Tangut a vs a̱. A common distinction is front /a ~ æ/ vs back /ɑ/ (e.g. Sūzhōu Wu 擦 tshaeh vs 尺 tshaoh; certain French varieties patte vs pâte), which can map to a and a̱ respectively. Another is higher/relaxed /ɐ/ vs lower/‘proper’ /ɑ ~ a/ (e.g. Hindi-Urdu a vs ā). English speakers can use both options:

  • Plain a [ɐ]: Like the vowel in cat [æ ~ a] or but [ɐ].
  • Uvularized a̱ [ɑʶ]: Like General American hot [ɑ], or RP/Southern English hard [aː ~ ɑː].

For those most comfortable with a language that does not have a robust distinction between two a-like sounds, such as Standard Mandarin:

  • Plain a: Use the vowel of English but or cat.
  • Uvularized a̱: Use your native a-sound, like Mandarin 坝 bà.

Vowels e [e] and e̱ [ɛ̠ʶ]

  • e [e]: High-mid front unrounded vowel. May have a palatal onglide [je].
  • e̱ [ɛ̠ʶ]: Uvularized low-mid (near-)front unrounded vowel.
e e̱
𗤋 me¹ [me˥˨] ‘not exist’ 𗑉 me̱¹ [mɛ̠ʶ˥˨] ‘eye’
𘟀 le² [le˧˧˦] ‘to see’ 𘜞 tse̱² [tsɛ̠ʶ˧˧˦] ‘for a while’

For those whose native language distinguishes between mid-high /e/ and mid-low /ɛ/ vowels, this distinction can be naturally applied to differentiate between e and e̱: French bébé vs bête, Standard German geh vs Bett. The uvularized e̱ may be further emphasized with actual uvularization and slight centralization.

Introducing an onglide can help exaggerate the difference between e and e̱:

  • The plain e should be articulated with a palatal glide [je], as in Russian пень or Mandarin 别 bié.
  • The uvularized e̱ should be articulated with a velar onglide [ɰɛ̠ʶ], similar to an over-articulated Russian мэр.

For monolingual speakers of Standard Mandarin, it is recommended to use 别 bié for e and mimic a monophthongized pronunciation of 白 bai for e̱.

Vowels i [i] and i̱ [ɪʶ]

  • Plain i [i]: High front unrounded vowel. May have a palatal onglide [ji].
  • Uvularized i̱ [ɪʶ]: Uvularized near-high (near-)front unrounded vowel.
Plain i Uvularized i̱
𘃡 vi¹ [ʋʲi˥˨] ‘to do’ 𗨻 vwi̱² [wɪʶ˧˧˦] ‘to become’
𗤢 bi² [bi˧˧˦] ‘low’ 𗾔 bi̱² [bɪʶ˧˧˦] ‘sun’
𗭪 si² [si˧˧˦] particle 𗁸 si̱² [sɪʶ˧˧˦] ‘tight’

Plain i, especially when exaggerated for distinction, can be pronounced with a palatal onglide [ji] or almost as a syllabic yod [j̩.]. Those who speak a language with this tendency (e.g., Mandarin 蜜 mì; French vie) can retain it when pronouncing Tangut.

The pedagogical recommendation for uvularized i̱ involves applying velarization and retraction to [i], potentially resulting in a sound approaching [ɰɪʶ] Think of an exaggerated diphthongal Russian ты (avoid, however, the monophthongal Polish ty) or the Turkish kıyma.

For monolingual speakers of Standard Mandarin:

  • Plain i: Use the Mandarin i final, as in 笔 bǐ.
  • Uvularized i̱: Mimic the vowel sound in Northern/Northwestern Mandarin pronunciation of ei, close to [ɪ], as in 北 běi, but pronounce it as a monophthong [pɪ].

Vowels o [ɵ] and o̱ [ɔʶ]

  • Plain o [ɵ]: High-mid central rounded vowel.
  • Uvularized o̱ [ɔʶ]: Uvularized open-mid back rounded vowel.
Plain o Uvularized o̱
𘊆 bo² [bɵ˧˧˦] ‘punishment’ 𘊳 bo̱¹ [bɔʶ˥˨] ‘forest’
𘝶𗦜 the²so² [tʰe.sɵ] ‘how’ 𗴀 so̱² [sɔʶ˧˧˦] ‘high’
𗼕 lo¹ [lɵ˥˨] ‘blessing’ 𗓽 lo̱¹ [lɔʶ˥˨] transcr.

The canonical reference for vowel quality is Mongolian ⟨ө/ᠥ⟩ vs ⟨о/ᠣ⟩. Kazakh ⟨ө/ٶ⟩ vs ⟨о/و⟩ can also serve as a reference, but avoid the diphthongization often present in Kazakh.

Although there is no harm in pronouncing o as ö [ø], it is preferable to front only slightly, targeting instead a central vowel [ɵ], i.e. midway between French peau and peu. The sound can also be achieved by starting with a schwa and adding rounding. For British English speakers:

  • Plain o [ɵ]: Similar to Southern English so, but without diphthongization.
  • Uvularized o̱ [ɔʶ]: Similar to RP or Southern English saw.

Vowels u [ʉ] and u̱ [ʊʶ]

  • Plain u [ʉ]: High central rounded vowel.
  • Uvularized u̱ [ʊʶ]: Uvularized near-high back rounded vowel.
Plain u Uvularized u̱
𗨁 phu² [pʰʉ˧˧˦] ‘up’ 𘕰 phu̱² [pʰʊʶ˧˧˦] ‘tree’
𗯴 khu¹ [kʰʉ˥˨] ‘below’ 𗼒 qhu̱¹ [qʰʊʶ˥˨] ‘to offer’
𘟣 du¹ [dʉ˥˨] ‘to exist’ 𗹶 du̱¹ [dʊʶ˥˨] ‘storeyed building’

The canonical reference for vowel quality is Kazakh ⟨ү/ٷ⟩ vs ⟨ұ/ۇ⟩, or Mongolian ⟨ү/ᠦ⟩ vs ⟨у/ᠤ⟩.

Although there is no harm in pronouncing u as ü [y], it is preferable to target instead a high central [ʉ], midway between French lu and loup. For British English speakers:

  • Plain u [ʉ]: Similar to Southern English too, but without diphthongization.
  • Uvularized u̱ [ʊʶ]: Similar to Southern English pull.

Vowels ə [ɨ] and ə̱ [ʌʶ]

  • Plain ə [ɨ]: Any high to high-mid, near-front to back, unrounded vowel. The actual quality could vary widely within the range [ɪ ~ ɨ ~ ɯ].
  • Uvularized ə̱ [ʌʶ]: Uvularized low-mid back unrounded vowel.
Plain ə Uvularized ə̱
𘉑 mə¹ [mɨ˥˨] ‘someone else’ 𗹦 mə̱¹ [mʌʶ˥˨] ‘heaven’
𗈶 sə¹ [sɨ˥˨] ‘to die’ 𗬐 sə̱¹ [sʌʶ˥˨] transcr.
𘙌 kə¹ [kɨ˥˨] affix 𗛖𗚝 qə̱¹dzi̱² [qʌʶ.dzɪʶ] ‘thorn’

The canonical reference for vowel quality is Kazakh ⟨і/ئ⟩ vs ⟨ы/ى⟩. In practice, a wide range of sounds can be used to approximate the Tangut vowel pair:

  • Plain ə [ɨ]: Any close-ish schwa-like sound sufficiently distinct from both [i] and [u], e.g. the vowel in Mandarin 丝 sī, Polish ty, Turkish alması, or the bit vowel in most English varieties.
  • Uvularized ə̱ [ʌʶ]: Preferably a mid-low and very back vowel, like [ɔ] but unrounded, similar to Korean ㅓ or very conservative RP but. Alternatively, an “energetic” schwa-like sound, such as British English hurt or Mandarin 色 sè is acceptable.

Diphthong uo [ʉ͡ɵ]

As a language with a tendency towards CV structure, Tangut generally lacks diphthongs as syllable nuclei. However, the GX202411 system recognizes one diphthong, uo [ʉ͡ɵ], corresponding to rhyme R.59 (= 1.57 𘉐 yuo¹) and possibly also R.60 (= 2.50 𗷢 tśuo²). It occurs only in non-uvularized (Grade III) contexts.

Plain uo
𘔼 nuo¹ [nʉ͡ɵ˥˨] ‘cause’
𘉐 yuo¹ [jʉ͡ɵ˥˨] ‘achievement’
𗔉 tśhuo¹ [tɕʰʉ͡ɵ˥˨] ‘perhaps, or’

Initial Consonants

Tangut philologists, following the Chinese tradition of 等韻學 děngyùnxué, categorized initial consonants into nine distinct classes. Within each class, a further distinction is made between what the author refers to as Sinoid initials (類漢聲母), considered by Tangut scholars to be analogous to initials in the Chinese variety they were familiar with, and epichoric initials (國音聲母), considered unique to the Tangut language. In the Sea of Characters, characters with Sinoid initials appear predominantly in the main rhyme-based chapters, whereas those with epichoric initials are found mostly in a separate volume, the Miscellaneous Categories.

A key feature of the GX202411 system is its systematic representation of uvularization-conditioned allophony. This means that certain initials are reconstructed with one pronunciation in plain syllables (Grade III) and a different pronunciation in uvularized syllables (Grade I/II). Such distinctions are sometimes phonetically obvious and sometimes specific to the Tangut language, but they are consistently represented in the transcription. These cases are always presented with the plain variant first, followed by the uvularized variant: v/vw-, k/q-, kh/qh-, g/ġ-, tś/tṣ-, tśh/tṣh-, ś/ṣ-, dź/dẓ-, gh/ġh-, ź/ẓ-.

Class Native designation Sinoid initials Epichoric initials
I-P 𗠉𗙏𗊢rmə̱r¹ġhai̱h² nllə¹ = Heavy labials p, ph, b, m —
II-V 𗠉𗙏𗥫rmə̱r¹ġhai̱h² ye¹ = Light labials v/vw, f —
III-T 𗢯𘁙𗙏lhwa¹dźwa¹ġhai̱h² = Dentals t, th, d, n —
IV-Ṇ 𗢯𗀔𗙏lhwa¹tśhaw¹ġhai̱h² = Cerebrals ṇ —
V-K 𘟗𗙏rqo̱ṃr²ġhai̱h² = Molars k/q, kh/qh, g/ġ, ŋ —
VI-S 𘘄𘁙𗙏świ¹dźwa¹ġhai̱h² = Sibilants ts, tsh, s dz
VII-Š 𘘄𗒘𗙏świ¹ġhae̱¹ġhai̱h² = Post-alveolars tś/tṣ, tśh/tṣh, ś/ṣ dź/dẓ
VIII-H 𗍂𗙏nwe¹ġhai̱h² = Gutturals h, gh/ġh, y, w gh/ġh, w, h, y Ø
IX-L 𘚶𘞨𗙏lə¹źeh²ġhai̱h² = Laterals/rhotics l, ll, r, z, ź/ẓ lh

Class I-P: 𗠉𗙏𗊢 rmə̱r¹ġhai̱h² nllə¹ = Heavy labials

Initial Plain (Grade III) Uvularized (Grade I)
p- 𗟱𗾞 pə¹nnə² [pɨ.nnɨ] ‘today’ 𘏞𘛛 po̱¹teṃ¹ [pɔʶ.tew̃] ‘bodhi’
ph- 𗉣 phi¹ [pʰi˥˨] ‘will, intent’ 𘕰 phu̱² [pʰʊʶ˧˧˦] ‘tree’
b- 𘞆 bi¹ [bi˥˨] ‘thin’ 𘊳 bo̱¹ [bɔʶ˥˨] ‘forest’
m- 𗴺 ma¹ [mɐ˥˨] ‘mother’ 𗑉 me̱¹ [mɛ̠ʶ˥˨] ‘eye’

Class II-V: 𗠉𗙏𗥫 rmə̱r¹ġhai̱h² ye¹ = Light labials

  • f- [f] or [ɸ]: Rare, occurs only in plain syllables (Grade III).

Uvularization-conditioned allophony:

  • v- [ʋ] or [β̞] (plain, Grade III)
  • vw- [β̞ʷ] or [w] with automatic rounding (uvularized, Grade I/II)
Initial Plain (Grade III) Uvularized (Grade I)
v/vw- 𘃡 vi¹ [ʋʲi˥˨] ‘to do’ 𘕣 vwa̱² [β̞ʷɑʶ˧˧˦] ‘what?’
f- 𘋇 fi¹ [fi˥˨] ‘mouse’

Class III-T: 𗢯𘁙𗙏 lhwa¹dźwa¹ġhai̱h² = Dentals

Initial Plain (Grade III) Uvularized (Grade I)
t- 𗓱 te¹ [te˥˨] ‘if’ 𗧊 to̱² [tɔʶ˧˧˦] ‘to come out’
th- 𗋕 tha¹ [tʰɐ˥˨] ‘that’ 𗢳 tha̱¹ [tʰɑʶ˥˨] ‘Buddha’
d- 𘟣 du¹ [dʉ˥˨] ‘to exist’ 𘋩 do̱² [dɔʶ˧˧˦] ‘place’
n- 𘟙 ne² [ne˧˧˦] ‘king’ 𗓰 na̱¹ [nɑʶ˥˨] ‘deep’

Class IV-Ṇ: 𗢯𗀔𗙏 lhwa¹tśhaw¹ġhai̱h² = Cerebrals

  • ṇ- [ɳʵ]: Retroflex nasal. Can be approximated as nr- or simply as n-. Note that, as a retroflex initial, it automatically triggers the insertion of the Grade II medial in uvularized contexts, resulting in Grade II syllables.
Initial Plain (Grade III) Uvularized (epenth. -a-, Grade II)
ṇ- 𘊄 ṇoṃ² [ɳʵɵw̃˧˧˦] ‘bad’ 𘝣 ṇae̱ṃ¹ [ɳʵa̯ˤɛw̃˥˨] ‘muddy’

Class V-K: 𘟗𗙏 rqo̱ṃr²ġhai̱h² = Molars

  • ŋ- [ɴ]: Occurs only in uvularized syllables (Grade I/II).

Uvularization-conditioned allophony:

  • k- [k], kh- [kʰ], g- [ɡ] velar (plain, Grade III)
  • q- [q], qh- [qʰ], ġ- [ɢ] uvular (uvularized, Grade I/II)
Initial Plain (Grade III) Uvularized (Grade I)
k/q- 𗑱 ka² [kɐ˧˧˦] ‘aeon’ 𗟔 qo̱¹ [qɔʶ˥˨] ‘carriage’
kh/qh- 𗯴 khu¹ [kʰʉ˥˨] ‘below’ 𘂤 qha̱¹ [qʰɑʶ˥˨] ‘among’
g/ġ- 𘒏 ga¹ [ɡɐ˥˨] ‘army’ 𗦬 ġu̱² [ɢʊʶ˧˧˦] ‘together’
ŋ- 𗧓 ŋa̱² [ɴɑʶ˧˧˦] ‘I’

Class VI-S: 𘘄𘁙𗙏świ¹dźwa¹ġhai̱h² = Sibilants

  • dz- [dz]: Epichoric initial.
  • z- [z] phonetically belongs here but is classified under Laterals/rhotics by Tangut phonologists.
Initial Plain (Grade III) Uvularized (Grade I)
ts- 𗥤 tse² [tse˧˧˦] ‘understand’ 𗥣 tsu̱² [tsʊʶ˧˧˦] ‘jump out’
tsh- 𗱱 tsha² [tsʰɐ˧˧˦] ‘repay’ 𗷖 tshwi̱w¹ [tsʰwəʶɰ˥˨] ‘direction’
s- 𗜍 sa¹ [sɐ˥˨] ‘kill’ 𗑗 se̱¹ [sɛ̠ʶ˥˨] ‘clean’
dz- 𗡅 dzi¹ [dzi˥˨] ‘eat’ 𗦢 dza̱² [dzɑʶ˧˧˦] ‘measure’
(z-) 𘄄 zi¹ [zi˥˨] ‘son’ 𗜈 zo̱ṃ² [zɔʶw̃˧˧˦] ‘hold’

Class VII-Š: 𘘄𗒘𗙏 świ¹ġhae̱¹ġhai̱h² = Post-alveolars (‘Shibilants’)

  • dź/dẓ-: Epichoric initial.
  • ź/ẓ- phonetically belongs here but is classified under Laterals/rhotics by Tangut phonologists.

Conditioned allophony triggered by uvularization:

  • tś- [tɕ], tśh- [tɕʰ], ś- [ɕ], dź- [dʑ] alveolo-palatal (plain, Grade III)
  • tṣ- [tʂ], tṣh- [tʂʰ], ṣ- [ʂ], dẓ- [dʐ] retroflex (uvularized, triggers epenthesis of Grade II medial, hence Grade II)
Initial Plain (Grade III) Uvularized (epenth. -a-, Grade II)
tś/tṣ- 𗵘 tśa¹ [tɕɐ˥˨] ‘road’ 𗉋 tṣao̱ṃ¹ [tʂa̯ˤɔw̃˥˨] ‘assemble’
tśh/tṣh- 𗉘 tśhə¹ [tɕʰɨ˥˨] ‘this’ 𗽇𗅾 tṣhaa̱¹no̱¹ [tʂha̯ˤa.nɔʶ] ‘moment’
ś/ṣ- 𗀓 śe² [ɕe˧˧˦] ‘to ask for’ 𗀚 ṣao̱¹ [ʂa̯ˤɔ˥˨] ‘to lead’
dź/dẓ- 𗱠 dźe¹ [dʑe˥˨] ‘pure, only’ 𘓟 dẓae̱² [dʐa̯ˤɛ˧˧˦] ‘to believe’
(ź/ẓ) 𗋾 źu² [ʑʉ˧˧˦] ‘fish’ 𗖣 rẓai̱r¹ [ʐɜ̯ˤɪ˞˥˨] ‘to scold’

Class VIII-H: 𗍂𗙏 nwe¹ġhai̱h² = Gutturals

  • gh/ġh- [ɣ] / [ʁ] voiced fricatives, with velar/uvular allophony conditioned by vowel uvularization, akin to Class V-K Velars.
  • w- [ɦw] a weaker version of ġh-, possibly close to zero initial, occurring only before uvularized (Grade I) vowels o̱ and u̱.
  • h- [h] glottal fricative, uncommon in native words. It can also be reconstructed as velar/uvular [x] / [χ]
  • y- [j] occurs only in plain syllables (Grade III).
  • Ø [ʔ] glottal stop, epichoric initial; occurs only word-initially in polysyllabic words.
Initial Plain (Grade III) Uvularized (Grade I)
gh/ġh- 𗐜 gha² [ɣɐ˧˧˦] ‘cover’ 𗼜 ġhai̱² [ʁˤɜ̯ˤɪ˧˧˦] ‘benefit’
w- 𘃽 wo̱² [ɦwɔʶ˧˧˦] ‘enter’
h- 𘜤 ha¹ [hɐ˥˨] ‘quickly’ 𗥑𗕌 hu̱¹źiṃ¹ [hʊʶ.ʑiw̃] ‘lady’
y- 𘝵 ye¹ [je˥˨] ‘self’
Ø 𗈪𗤶 a¹nne¹ [ʔɐ.nne] ‘singly’ 𗠝𗗆𗒶 a̱¹si̱ṃ¹khi¹ [ʔɑʶ.sɪʶw̃.kʰi] ‘asaṃkhyeya’

Class IX-L: 𘚶𘞨𗙏 lə¹źeh²ġhai̱h² = Laterals and rhotics

  • l- [l] cross-linguistically common lateral approximant. Can be pronounced as velarized ‘dark L’ [ɫ] in uvularized (Grade I) contexts if this feels more natural.
  • ll- [ɮ] possibly with stop release [ɮᵈ]; a ‘hard’ lateral fricative, like Khalkha Mongolian ⟨л⟩. Pedagogically, it can be simplified to l-, pronounced as ld- or as a long ll-. Pay attention to avoid pronouncing ll- like Welsh or Catalan ⟨ll⟩.
  • r- [ɻ] retroflex approximant, found almost exclusively with rhotacized vowels. English initial r- but without lip rounding; Mandarin r- but with less frication.
  • z- [z] see Class VI-S Sibilants.
  • ź/ẓ- [ʑ] / [ʐ] see Class VII-Š Post-alveolars; note the uvularization-conditioned allophony.
  • lh- [ɬ] possibly with stop release [ɬᵗʰ]; voiceless lateral fricative, like Welsh ⟨ll⟩. Similar to Tibetan ⟨lh⟩, but Tibetan ⟨lh⟩ is often a ‘soft’ devoiced approximant [l̥], whereas Tangut lh- is likely a ‘hard’ voiceless fricative [ɬ]
Initial Plain (Grade III) Uvularized (Grade I, ẓ- Grade II)
l- 𘟀 le² [le˧˧˦] ‘see’ 𘈩 li̱w¹ [ləʶɰ˥˨] ‘one’
ll- 𘜶 lleh² [ɮḛ˧ˀ] ‘big’ 𗱈 lli̱² [ɮɪʶ˧˧˦] ‘tiger’
r- 𘆝 rer¹ [ɻe˞˥˨] ‘horse’ 𗦎 ra̱r² [ɻɑ˞ʶ˧˧˦] ‘pass, leave’
z- 𘄄 zi¹ [zi˥˨] ‘son’ 𗜈 zo̱ṃ² [zɔʶw̃˧˧˦] ‘hold’
ź/ẓ- 𗋾 źu² [ʑʉ˧˧˦] ‘fish’ 𗖣 rẓai̱r¹ [ʐɜ̯ˤɪ˞˥˨] ‘scold’
lh- 𗽀 lhə² [ɬɨ˧˧˦] ‘dust’ 𗇘 lhi̱w² [ɬəʶɰ˧˧˦] ‘release’

Pan-Syllabic Features

Tense Vowels: -h

The First minor cycle of rhymes (R.61–R.76 according to Gong Hwang-cherng) are widely recognized as sharing a distinctive phonological feature, conventionally termed tense. Gong Hwang-cherng marks these with a subscript dot (e.g. ạ), while Arakawa uses a final -q. This transcription system indicates such syllables with an orthographic -h placed after the syllable nucleus and any potential coda, e.g. 𗅲 teṃh².

The prevailing interpretation assigns to such ‘tense’ syllables a checked or glottalized quality, similar to the Chinese entering tone in dialects with glottal-type reflexes, or the Danish stød. In the absence of a more compelling alternative, this system adheres to the consensus view. According to this view, tense syllables exhibit:

  • Phonation ranging from stiff to creaky, potentially culminating in a weak glottal stop [-ʔ] at the syllable end.
  • When word-final, the vowel tends to be shorter and abruptly cut off.
Tense Vowel Comparison
𗏡 qu̱h¹ [qʊ̰ʶ˥˧ˀ] ‘later’ 𗌭 qu̱¹ [qʊʶ˥˨] ‘if’
𘎞 na̱h¹ [nɑ̰ʶ˥˧ˀ] ‘nasal mucus’ 𗓰 na̱¹ [nɑʶ˥˨] ‘deep’
𗦁 ġhai̱h² [ʁˤɜ̯ˤɪ̰˧ˀ] ‘obstacle’ 𗼜 ġhai̱² [ʁˤɜ̯ˤɪ˧˧˦] ‘benefit’
𗉝 śih² [ɕḭ˧ˀ] ‘grass’ 𗶹 śi² [ɕi˧˧˦] ‘to go’
  • The exact boundary between the first and second minor cycles remains debated. Arakawa, for instance, places the end of the first minor cycle at R.79, thus reconstructing R.77–R.79 with tense vowels (-yeq2, -eq'2, -yeq'2). The final ‘wastebasket’ rhyme chapters likely also include some tense rhymes. For R.104, sources within the GHC tradition vary between non-tense -ũ and tense -ụ̃. Arakawa reconstructs R.102 as -woq2.
  • The -h notation used here is intentionally non-committal regarding the precise phonetic nature of ‘tense’ syllables, allowing for potential future reinterpretations. If the consensus holds, -h should be read as the common orthographic device in other romanization systems for East Asian languages (e.g. POJ/KIP for Nuclear Southern Min = Hokkien/Taiwanese, 吳語協會 Wúyǔ Xiéhuì romanization for Wu) to indicate a final glottal stop or glottalization, e.g. 白 as POJ pe̍h and Shanghainese (Wúyǔ Xiéhuì) bah.

Retroflex Vowels: r-...-r

Syllables in the Second Minor Cycle (R.77–R.98 according to Gong Hwang-cherng) are generally believed to exhibit syllable-wide retroflexion, often referred to as having retroflex or rhotacized vowels. Both GHC and Arakawa typically denote this feature with a single final -r. This transcription system uses a bracketing r-...-r notation. The following orthographic rules apply:

  • If the initial consonant is r-, which is compatible only with retroflex rhymes, the pre-initial r- is omitted, e.g. 𘜘 rir¹ ‘to obtain’ (not *rrir¹).
  • The suffix -r is placed after any potential coda, e.g. 𗄑 rŋo̱ṃr² ‘all’ (not *rŋo̱rṃ²).

This bracketing notation is placed at the outermost edges of the syllable to suggest a holistic ‘Gestalt’ recognition, indicating an r-colouring applied throughout the syllable. The most likely phonetic realization involves true apical retroflexion of the vowel [V˞], which can be difficult to perceive and produce for those who do not speak a language with such vowels (rhotic English and Mandarin do not count). As a pedagogical convention, this quality can be reinforced by adding an r-sound (any non-guttural type) after the vowel or syllable.

Common non-uvularized (Grade III) rhymes with retroflex vowels include:

Rhyme Example Comparison
-ar 𘉋 ryar¹ [ʵʝɐ˞˥˨] ‘eight’ 𘁂 ya²
-er 𗇋 rmer² [me˞˧˧˦] (nom.) 𗷝 me² negator
-ir 𗧹 rgir¹ [ɡi˞˥˨] ‘disaster’ 𗊞 gi¹ ‘awake’
-or 𗌮 rmor¹ [mɵ˞˥˨] ‘real’ 𗔘 mo¹ ‘hear’
-ur 𗑯 rgur² [ɡʉ˞˧˧˦] ‘lie down’ 𗅧 gu² ‘tendon’
-ər 𘐴 rgər² [ɡɨ˞˧˧˦] ‘beat’ 𘈚 gə² ‘night’

Common uvularized (Grade I/II) rhymes with retroflex vowels include:

Rhyme Example Comparison
-a̱r 𗏘 rġha̱r² [ʁɑ˞ʶ˧˧˦] ‘chest’ 𘕿 ġha̱² loc. particle
-e̱r 𗮅 re̱r² [ɻɛ̠˞ʶ˧˧˦] ‘many’ 𗑉 me̱¹ ‘eye’
-i̱r 𗄛 rbi̱r² [bɪ˞ʶ˧˧˦] ‘meet’ 𗾔 bi̱² ‘sun’
-o̱r 𘑠 rvwo̱r² [βʷɔ˞ʶ˧˧˦] ‘calf’ 𗧘 vwo̱² ‘meaning’
-ai̱r 𘇒 rvwai̱r¹ [βʷɜ̯ˤɪ˞˥˨] ‘love’ 𗪺 ġhai̱¹ ‘power’
-ao̱r 𗩽 rtṣao̱r² [tʂa̯ˤɔ˞˧˧˦] ‘filth’ 𗀚 ṣao̱¹ ‘lead’
-u̱r 𗏨 rġhu̱r¹ [ʁʊ˞ʶ˥˨] ‘corpse’ 𗥦 ġhu̱¹ ‘head’
-ə̱r 𗰜 rmə̱r² [mʌ˞ʶ˥˨] ‘root’ 𘋠 mə̱² ‘kind’
  • The r-...-r bracketing deviates from the common practice of using a single final -r, so some brief justification is in order. In addition to indicating the whole-syllable nature of rhotacization, this approach aligns with the principle of keeping our transcription of Tangut close to medieval transcriptions of Tangut from and into other languages, especially those with a standard modern romanization. Tibetan transcriptions of Tangut sometimes use a pre-initial r- and sometimes a final -r for rhotacized syllables:
    • 𘉋 ryar¹: ⟨rye⟩
    • 𗹙 rtser¹: ⟨rtse⟩
    • 𗬫 rdər²: ⟨'dar⟩, ⟨dhir⟩

Segmental Features

Nasal Preinitial: n-, m-, ŋ- /N-/

Rhyme classes known as the ‘parallel columns’ are represented in GHC as long vowels (aa or, confusingly, a̱) and in Arakawa as a purely notational prime ('). Following Gong (2021), these are reinterpreted as distinguished by a segmental nasal preinitial /N-/. This system represents this pre-initial orthographically based on the subsequent initial consonant:

  • Before stops/affricates, it is written as a homorganic nasal: nt-, nd-, ntsh-, ndz-, ntś-, ndź-; mph-, mb-; ŋkh-, ŋg-, ŋqh-, ŋġ-. Examples: 𗡢 nto¹, 𗶠 ndzu̱², 𘜼 mpho², 𗥞 ŋgu²
  • Before nasals, it is written as a geminate nasal: mm-, nn-, ŋŋ-. Examples: 𘈞 mme¹, 𗤶 nne¹, 𗟲 ŋŋwu̱¹
  • Before fricatives/approximants, it is consistently written as n-: nv-, ns-, nz-, nś-, nṣ-, ny-. Examples: 𗈁 nvu¹, 𗾫 nse², 𗧯 nye¹

Examples before stops/affricates (homorganic nasal):

Initial Cluster Example
nt- 𗡢 nto¹ [ntɵ˥˨] ‘to seek’
nth- 𗤓 ntho¹ [ntʰɵ˥˨] ‘wonderful’
nd- 𗘳 ndu² [ndʉ˧˧˦] ‘to stab’
ntsh- 𘎪 ntshe¹ [ntsʰe˥˨] ‘to say’
ndz- 𗶠 ndzu̱² [ndzʊʶ˧˧˦] ‘to sit’
ntś- 𗈙 ntśu¹ [ntɕʉ˥˨] ‘harm, slander’
ndź- 𗫻 ndźe¹ [ndʑe˥˨] ‘to dwell’
mph- 𘜼 mpho² [mpʰɵ˧˧˦] ‘to combine’
mb- 𗠁 mbu̱² [mbʊʶ˧˧˦] ‘victory’
ŋkh- 𗪀 ŋkhu² [ŋkʰʉ˧˧˦] ‘to see’
ŋg- 𗥞 ŋgu² [ŋɡʉ˧˧˦] ‘to save’
ŋqh- 𗉫 ŋqhe̱¹ [ɴqʰɛ̠ʶ˥˨] ‘to enjoy’
ŋġ- 𘆕 ŋġa̱² [ɴɢɑʶ˧˧˦] ‘thin, emaciated’

Examples before nasals (geminate nasal):

Initial Cluster Example Comparison
mm- 𘈞 mme¹ [mme˥˨] ‘dream’ 𗤋 me¹ ‘not’
nn- 𗤶 nne¹ [nne˥˨] ‘heart’ 𗶚 ne¹ ‘near’
ŋŋ- 𗟲 ŋŋwu̱¹ [ɴɴwʊʶ˥˨] ‘speech’ 𘟂 ŋwu̱² ‘is’

Examples before fricatives/approximants (written n-):

Initial Cluster Example
nv- 𗈁 nvu¹ [n̚ʋʉ˥˨] ‘mercy’
ns- 𗾫 nse² [n̚se˧˧˦] ‘thought’
nz- 𘒮 nzi̱w² [n̚zəʶɰ˧˧˦] ‘endure’
nś- 𘝇 nświ² [n̚ɕwi˧˧˦] ‘harmony’
nṣ- 𗰖 nṣao̱¹ [n̚ʂa̯ˤɔ˥˨] ‘collect’
ny- 𗧯 nye¹ [n̚ʝe˥˨] ‘hold, maintain’

Round Medial: -w- /-w-/

Tangut syllables are distinguished by the presence vs absence of lip rounding, dividing them into ‘round’ (合口 hékǒu) and ‘non-round’ (開口 kāikǒu) types. Most reconstruction systems, including this one, analyse round syllables as having a distinct medial element /-w-/ between the initial and the nucleus, transcribed here as _-w_.

Round Syllable Comparison
𗒱 świ¹ [ɕwi˥˨] ‘age’ 𗪘 śi¹ ‘before’
𗰸 khwə¹ [kʰwɨ˥˨] ‘cut off’ 𗭒 khə¹ ‘foot’

Grade II ‘ʕayn’ Medial: -a- /-ʕ-/

Grade II syllables are analysed as phonologically uvularized (Gong 2020), similar to Grade I, but differ by the presence of an additional ‘Grade II medial’. GHC represents this medial as -i-, while Arakawa uses -y-. This system diverges, positing an a-like medial, likely a pharyngeal fricative/approximant /ʕ/ (like Arabic ع ʕayn), hence the term ‘ʕayn medial’. Orthographically, it is represented simply as -a- preceding the main vowel nucleus (uvularized, hence horizontal line below): aa̱, ae̱, ai̱, ao̱, au̱, aə̱.

GX IPA Approximation Example
aa̱ [a̯ˤa] aa, ea 𗞽 baa̱¹ [ba̯ˤa˥˨] ‘burst’
ae̱ [a̯ˤɛ] ae, ai 𗒘 ġhae̱¹ [ʁˤa̯ˤɛ˥˨] ‘true’
ai̱ [ɜ̯ˤɪ] ei, əi 𗪺 ġhai̱¹ [ʁˤɜ̯ˤɪ˥˨] ‘power’
ao̱ [a̯ˤɔ] ao, eo 𗀋 phao̱² [pʰa̯ˤɔ˧˧˦] ‘snake’
au̱ [ɜ̯ˤʊ] ou, eu 𘔯 ŋqhau̱¹ [ɴqˤʰɜ̯ˤʊ˥˨] ‘greet’
aə̱ [ɜ̯ˤə] aə, eə 𘇹 ṣaə̱¹ [ʂɜ̯ˤə˥˨] ‘thread’

In specific phonological contexts, this Grade II medial -a- appears automatically (epenthetically) in all uvularized contexts:

  1. Uvular before i̱: Uvular consonants (Class V-K: q-, qh-, ġ-, Class VIII-H: ġh-) cannot directly precede the uvularized vowel i̱. An -a- is obligatorily inserted, yielding Qai̱ sequences like 𗣓 ġai̱¹ ‘difficult’ (not *ġi̱¹). Other examples include 𗎍 qhai̱¹ ‘hate’, 𗵒 qai̱h¹ ‘gold’, 𗙏 ġhai̱h² ‘sound’.
  2. After retroflex initials: Retroflex initials (Class IV-Ṇ: ṇ-, Class VII-Š: tṣ-, tṣh-, dẓ-, ṣ-; Class IX-L: ẓ-) cannot directly precede any uvularized vowel; the Grade II medial -a- is inserted without exception. Examples: 𘇹 ṣaə̱¹ ‘thread’, 𘓟 dẓae̱² ‘to believe’, 𗉋 tṣao̱ṃ¹ ‘to gather’, 𘝣 ṇae̱ṃ¹ ‘turbid’.

Yod Medial (Marginal): -y- /-j-/

Tangut Grade III syllables do not have a distinct yod medial. The high front vowels e and i may be pronounced with an onglide ([je], [ji]), but this is not what distinguishes them from their Grade I counterparts e̱ and i̱. This is most evident in the exception that confirms the rule: characters created by the Tanguts specifically to transcribe foreign syllables containing a /-j-/ medial. These characters are often fǎnqiè-derived characters (切音字 qièyīnzì), combining an initial speller with the character 𘁂 ya². Examples include:

  • 𘗔 dya² (derived from 𘗐 de² + 𘁂 ya²)
  • 𗓮 mya² (derived from 𗓁 mi¹ + 𘁂 ya²)

These characters are likely confined to the specialized language of Buddhist professionals, making the yod medial not only marginal to the language system but also peripheral to the langue of the language community.

Velar Coda: -w /-ɣ/

Rhymes R.44–49 in the major cycle are generally reconstructed with a w-like coda across different systems: GHC uses -w, Arakawa uses -u, and this system uses -w. Correspondences for other rhymes are complex. For example, GHC’s -ow for R.56–60 is interpreted differently here: R.56–58 as nasal codas (-o̱ṃ, -ao̱ṃ, -oṃ) and R.59 (tentatively also R.60) as diphthong -uo. Conversely, GHC’s -jaa (R.21) and -jwo (R.50) correspond to -aw and -ow in this system. Such complexities also exist outside the major cycle. Although written -w, this coda is considered a non-rounded velar /ɣ/, phonetically [-ɯ ~ -ɰ ~ -ɣ].

Rhyme Example
-aw 𗅔 naw² [nɐɰ˧˧˦] ‘be not’
-ow 𘘦 tśhow¹ [tɕʰow˥˨] ‘(only) then’
-i̱w 𘈩 li̱w¹ [ləʶɰ˥˨] ‘one’
-ai̱w 𗫦 ġhai̱w¹ [ʁˤəˤɰ˥˨] ‘learn’
-iw 𗤁 tśhiw¹ [tɕʰɪɰ˥˨] ‘six’

Nasal Coda: -ṃ /-w̃/

All major reconstruction systems recognize nasal features in certain rhymes. In the major cycle, rhymes R.15–16 and R.25–27 are universally acknowledged as nasal. GHC denotes these with tildes (e.g. ã, ĩ), while Arakawa uses a final -n. This system employs -ṃ. Elsewhere, correspondences are more complex; for example, major cycle rhymes R.41–43 and R.56–58, which GHC reconstructs as əj/ɨj and ow respectively, are considered nasal by both Arakawa and this system. This system interprets the nasality as asynchronous, meaning that instead of simultaneous vowel nasalization, there is a sequential coda segment, represented as -ṃ and likely pronounced as a nasalized labio-velar approximant [-w̃].

Common rhymes with a nasal coda, found in both loanwords and frequently in native words, include:

Rhyme Example
-e̱ṃ 𗣫 tse̱ṃ¹ [tsɛ̠ʶw̃˥˨] ‘small’
-ae̱ṃ 𘝣 ṇae̱ṃ¹ [ɳʵa̯ˤɛw̃˥˨] ‘turbid’
-eṃ 𗵆 śeṃ¹ [ɕew̃˥˨] ‘complete’
-o̱ṃ 𗜈 zo̱ṃ² [zɔʶw̃˧˧˦] ‘hold’
-ao̱ṃ 𗴂 phao̱ṃ¹ [pʰa̯ˤɔw̃˥˨] ‘white’
-oṃ 𘓯 khoṃ¹ [kʰɵw̃˥˨] ‘give’

Common rhymes with a nasal coda, found almost exclusively in Chinese loanwords (with 𘋢𗥤 dẓaa̱ṃ²tse² being one of the few exceptions), include:

Rhyme Example
-a̱ṃ 𘈬𗦺 nde²pha̱ṃ¹ [nde.pʰɑʶw̃] ‘nirvana’
-aa̱ṃ 𘋢𗥤 dẓaa̱ṃ²tse² [dʐa̯ˤaw̃.tse] ‘bodhisattva’
-aṃ 𗇁 śaṃ¹ [ɕɐw̃˥˨] ‘dhyāna’
-i̱ṃ 𗉺𗐺 ṣaa̱¹mi̱ṃ¹ [ʂa̯ˤa.mɪʶw̃] ‘śramaṇa’
-iṃ 𗥑𗕌 hu̱¹źiṃ¹ [hʊʶ.ʑiw̃] ‘lady, 夫人 fūrén’

List of rhyme chapters

Each rhyme chapter is provided with the following information:

  • R.x: Rhyme number modulo tone.
  • 1.xx: Tone 1 chapter number and representative syllable.
  • 2.xx: Tone 2 chapter number and representative syllable.
  • GHC: Gong Hwang-cherng’s transcription.
  • Ar.: Arakawa Shintarō’s transcription.
  • GX: GX202411 transcription.

The per-rhyme reconstructions in the GX202411 transcription system are provided for reference purposes. The values given are subject to change without triggering a version number increment in the transcription system. This is especially the case for the reconstructions currently marked as tentative, such as (-uo?) for R.60 = 2.50 𗷢 tśuo².

Major cycle (RR. 1–60)

Tone 1 Tone 2 GHC Ar. GX
R.1 1.01 𘁨 bu̱¹ 2.01 𗾴 bu̱² -u -u -u̱
R.2 1.02 𘜲 śu¹ 2.02 𗲺 śu² -ju -yu -u
R.3 1.03 𗔠 su¹ 2.03 𘆶 su² -ju -yu -u
R.4 1.04 𘙼 mu̱¹ 2.04 𘓩 mu̱² -u -u: -u̱
R.5 1.05 𘃠 ndu̱¹ 2.05 𗷪 ndu̱² -uu -u' n- -u̱
R.6 1.06 𗃴 ŋqhau̱¹ -juu -yu' n- -au̱
R.7 1.07 𗴢 mbu¹ 2.06 𗩈 mbu² -juu -u:' n- -u
R.8 1.08 𗬃 si̱¹ 2.07 𗙐 si̱² -e -i -i̱
R.9 1.09 𗠭 qai̱¹ 2.08 𗎃 qai̱² -ie -yi -ai̱
R.10 1.10 𗚋 tśi¹ 2.09 𗄐 tśi² -ji -i: -i
R.11 1.11 𗣣 tshi¹ 2.10 𘒇 tshi² -ji -i: -i
R.12 1.12 𗛁 mbi̱¹ 2.11 𗵿 mbi̱² -ee -i' n- -i̱
R.13 1.13 𘎛 nṣai̱¹ -iee -yi' n- -ai̱
R.14 1.14 𗮦 ntśi¹ 2.12 𘏗 ntśi² -jii -i:' n- -i
R.15 1.15 𗗆 si̱ṃ¹ 2.13 𗖗 si̱ṃ² -ẽ -in -i̱ṃ
R.16 1.16 𗑃 tśiṃ¹ -jĩ -yin -iṃ
R.17 1.17 𗊡 ha̱¹ 2.14 𗊺 ha̱² -a -a -a̱
R.18 1.18 𗉺 ṣaa̱¹ 2.15 𗺩 ṣaa̱² -ia -ya -aa̱
R.19 1.19 𗰭 śa¹ 2.16 𗜓 śa² -ja -a: -a
R.20 1.20 𘅄 sa¹ 2.17 𗾥 sa² -ja -a: -a
R.21 1.21 𘄢 yaw¹ 2.18 𘏧 yaw² -jaa -ya: -aw
R.22 1.22 𗴡 ŋġa̱¹ 2.19 𘆕 ŋġa̱² -aa -a' n- -a̱
R.23 2.20 𗫬 ntṣaa̱² -iaa -ya' n- -aa̱
R.24 1.23 𗫴 mma¹ 2.21 𘜴 mma² -jaa -a:' n- -a
R.25 1.24 𘂦 ha̱ṃ¹ 2.22 𘟑 ha̱ṃ² -ã -an -a̱ṃ
R.26 1.25 𘅻 ġhaa̱ṃ¹ 2.23 𗹺 ġhaa̱ṃ² -iã -yan -aa̱ṃ
R.27 1.26 𘓾 yaṃ¹ 2.24 𗶕 yaṃ² -jã -a:n -aṃ
R.28 1.27 𗢁 hə̱¹ 2.25 𗍉 hə̱² -ə -I -ə̱
R.29 1.28 𗙌 qaə̱¹ 2.26 𗞨 qaə̱² -iə -yI -aə̱
R.30 1.29 𘎻 śə¹ 2.27 𗶽 śə² -jɨ -I: -ə
R.31 1.30 𗝚 sə¹ 2.28 𗸹 sə² -jɨ -I: -ə
R.32 1.31 𘚾 ndə̱¹ -əə -I' n- -ə̱
R.33 1.32 𘃔 ndə¹ 2.29 𗠸 ndə² -jɨɨ -yI' n- -ə
R.34 1.33 𗃰 he̱¹ 2.30 𗨾 he̱² -ej -e -e̱
R.35 1.34 𗷘 qae̱¹ 2.31 𗷛 qae̱² -iej -ye -ae̱
R.36 1.35 𘟅 tśhe¹ 2.32 𗜙 tśe² -jij -e: -e
R.37 1.36 𘒋 tshe¹ 2.33 𗆎 tshe² -jij -e: -e
R.38 1.37 𘈧 nde̱¹ 2.34 𗥉 nde̱² -eej -e' n- -e̱
R.39 1.38 𘘅 ŋqae̱¹ -ieej -ye' n- -ae̱
R.40 1.39 𘅡 nse¹ 2.35 𗾫 nse² -jiij -e:' n- -e
R.41 1.40 𗣫 tse̱ṃ¹ -əj -e:' -e̱ṃ
R.42 1.41 𘐝 ṣae̱ṃ¹ 2.36 𗨆 ṣae̱ṃ² -iəj -en -ae̱ṃ
R.43 1.42 𗵆 śeṃ¹ 2.37 𗼃 śeṃ² -jɨj -yen -eṃ
R.44 1.43 𗺹 hi̱w¹ 2.38 𗬦 hi̱w² -ew -eu -i̱w
R.45 1.44 𗶥 qai̱w¹ 2.39 𗉡 qai̱w² -iew -yeu -ai̱w
R.46 1.45 𘝖 yiw¹ 2.40 𗺶 yiw² -jiw -eu: -iw
R.47 1.46 𗪲 ghiw¹ -jiw -eu: -iw
R.48 2.41 𗅢 nsi̱w² -eew -eu' n- -i̱w
R.49 1.47 𘞘 nyiw¹ -jiiw -yeu' n- -iw
R.50 1.48 𗄈 śow¹ -jwo -o -ow
R.51 1.49 𘀁 tho̱¹ 2.42 𗟌 thwo̱² -o -o -o̱
R.52 1.50 𗆉 qao̱¹ 2.43 𘋘 qao̱² -io -yo -ao̱
R.53 1.51 𗔆 so¹ 2.44 𗚂 so² -jo -o: -o
R.54 1.52 𗫕 ntho̱¹ 2.45 𘉼 ntho̱² -oo -o' n- -o̱
R.55 1.53 𗠟 nśo¹ 2.46 𘘁 nśo² -ioo, joo -yo' n- -ao̱, n- -o
R.56 1.54 𘙾 ho̱ṃ¹ 2.47 𗫫 ho̱ṃ² -ow -on -o̱ṃ
R.57 1.55 𘙠 qao̱ṃ¹ 2.48 𗽕 qao̱ṃ² -iow, -jow -yon -ao̱ṃ, -oṃ
R.58 1.56 𗭴 yoṃ¹ 2.49 𗭃 hoṃ² -jow -o:n -oṃ
R.59 1.57 𘉐 yuo¹ -ioow -o" -uo
R.60 2.50 𗷢 tśuo² -joow -yo" (-uo?)

First minor cycle (RR. 61–76?): Tense Vowels -h

Tone 1 Tone 2 GHC Ar. GX
R.61 1.58 𗚽 du̱h¹ 2.51 𗏒 du̱h² -ụ -uq -u̱h
R.62 1.59 𗤧 tśuh¹ 2.52 𘞏 lluh² -jụ -yuq -au̱h, -uh
R.63 1.60 𗧎 qae̱h¹ 2.53 𗐯 qae̱h² -iẹj -yeq -e̱h, -ae̱h
R.64 1.61 𗖞 tśeh¹ 2.54 𗿩 tśeh² -jịj -enq -eh
R.65 1.62 𗝾 teṃh¹ 2.55 𗅲 teṃh² -jɨ̣j -yenq -eṃh
R.66 1.63 𘟫 ġha̱h¹ 2.56 𗰗 ġha̱h² -ạ -aq -a̱h
R.67 1.64 𗕹 tah¹ 2.57 𘐓 tah² -jạ -a:q -aa̱h, -ah
R.68 1.65 𗄩 bi̱h¹ 2.58 𗇠 bi̱h² -ẹ -iq -i̱h
R.69 1.66 𗉁 ġhai̱h¹ 2.59 𗙑 ġhai̱h² -iẹ -yiq -ai̱h
R.70 1.67 𗳃 bih¹ 2.60 𗒣 bih² -jị -i:q -ih
R.71 1.68 𗨹 ġhə̱h¹ -ə̣ -iq' -ə̱h
R.72 1.69 𗥇 təh¹ 2.61 𗲸 təh² jɨ̣ -i:q' -aə̱h, -əh
R.73 1.70 𘔸 wo̱ṃh¹ 2.62 𗥏 wo̱ṃh² -ọ -oq -o̱ṃh
R.74 1.71 𗶪 to̱h¹ 2.63 𘒵 to̱h² -iọ -onq -o̱h
R.75 1.72 𗅼 toh¹ 2.64 𘈎 toh² -jọ -yonq -oh
R.76 2.65 𗽌 ġhae̱ṃh² -iə̣j -eq2 (-ae̱ṃh?)

Second minor cycle (RR. 77?–98): Rhotacized Vowels r-...-r

Tone 1 Tone 2 GHC Ar. GX
R.77 1.73 𗯿 rvwe̱r¹ 2.66 𗯬 rvwe̱r² -ejr -yeq2 r- -e̱r
R.78 2.67 𘌀 rġhae̱r² -iejr -eq'2 r- -ae̱r
R.79 1.74 𗙨 ryer¹ 2.68 𗳭 ryer² -jijr -yeq'2 r- -er
R.80 1.75 𗌈 rmu̱r¹ 2.69 𗎑 rmu̱r² -ur -ur r- -u̱r
R.81 1.76 𗋿 ryur¹ 2.70 𗌏 ryur² -jur -yur r- -ur
R.82 1.77 𗨈 rbi̱r¹ 2.71 𗄛 rbi̱r² -er -ir r- -i̱r
R.83 1.78 𗈮 rġhai̱r¹ -ier -yir r- -ai̱r
R.84 1.79 𗟶 ryir¹ 2.72 𘊝 ryir² -jir -i:r r- -ir
R.85 1.80 𗴐 rġha̱r¹ 2.73 𗐔 rġha̱r² -ar -ar r- -a̱r
R.86 1.81 𗉾 rṣaa̱r¹ -iar -yar r- -aa̱r
R.87 1.82 𗸪 ryar¹ 2.74 𗔻 ryar² -jar -a:r r- -ar
R.88 1.83 𗧵 nsa̱r¹ -aar -ar' n- -a̱r
R.89 2.75 𗡗 ryawr² -jaar -yar' r- -awr
R.90 1.84 𗙅 rhə̱r¹ 2.76 𘟐 rhə̱r² -ər -Ir r- -ə̱r
R.91 1.85 𗱘 rġhaə̱r¹ -iər -yIr r- -aə̱r
R.92 1.86 𗲆 rghər¹ 2.77 𗬼 ryər² -jɨr -I:r r- -ər
R.93 1.87 𗨎 rŋi̱wr¹ 2.78 𘜔 rŋi̱wr² -ewr -er r- -i̱wr
R.94 1.88 𗖬 rkiwr¹ 2.79 𘞫 rgiwr² -jiwr -yer r- -iwr
R.95 1.89 𘈆 rġho̱r¹ 2.80 𘏴 rġho̱r² -or -or r- -o̱r
R.96 1.90 𗑑 rmao̱r¹ 2.81 𗟽 rmor² -ior, jor -yor r- -ao̱r, r- -or
R.97 1.91 𗔶 rqo̱ṃr¹ 2.82 𗡹 rqo̱ṃr² -owr -o:r r- -o̱ṃr
R.98 2.83 𘊈 rnoṃr² -jowr -wor (r- -oṃr?)

Wastebasket group (RR. 99–105)

Most reconstructions assigned in previous systems are doubtful. My own reconstructions are also mostly tentative.

Tone 1 Tone 2 GHC Ar. GX
R.99 2.84 𗁌 nde̱r² -eer -ywor (n- -e̱r?)
R.100 1.92 𗂴 rtśər¹ 2.85 𗉕 rtśər² -jɨɨr -yIr (r- -ər?)
R.101 1.93 𗹙 rtser¹ 2.86 𗎫 rtser² -jiir -yer2 (r- -er?)
R.102 1.94 𗂌 ryor¹ -oor -woq2 (r- -or?)
R.103 1.95 𗌜 rnor¹ -joor -ya:n (r- -or?)
R.104 1.96 𗜰 tsu̱ṃ¹ -ũ -un -u̱ṃ
R.105 1.97 𗇜 rywar¹ -jwar -ua (r- -ar?)

Bibliography

  • Arakawa, Shintarō. 荒川慎太郎. 1999. 〈夏藏対音資料からみた西夏語の声調〉，《言語学研究》17-18:27-44.
  • Gong, Hwang-cherng. 2003. Tangut. In Randy LaPolla and Graham Thurgood (eds.), The Sino-Tibetan Languages, 602-620. London & New York: Routledge.
  • Gong, Xun. 2020. Uvulars and uvularization in Tangut phonology. Language and Linguistics 21.2: 175-212.
  • Gong, Xun. 2021. Nasal preinitials in Tangut phonology. Archiv orientální 89.3: 443-482.
  • Gong, Xun. 2025. Tone values of Tangut. Bulletin of Chinese Linguistics 18.1.
  • Nishida, Tatsuo. 西田龍雄. 1964. 《西夏語の研究 1》. 東京：座右宝刋行会.

Changelog

Only major changes are listed here. For the complete revision history, see the GitHub repository.

  • 2025-04-07: Initial public draft.