How to read the boxes in the tree. The deepest box (where known) is the
Old Chinese reconstruction, then the Middle Chinese Baxter–Sagart spelling;
early centuries keep that spelling, then each branch switches to pinyin-style (Mandarin) or
jyutping-style (Cantonese) spelling, and the newest boxes show standard
Hanyu Pinyin / Jyutping .
Tone categories 四聲
(none) · ¹ 平 píng — level tone
X · ² 上 shǎng — rising tone
H · ³ 去 qù — departing tone
-p -t -k · ⁴ 入 rù — entering (checked) tone, ends in a stop
In Baxter's notation X and H are written after the syllable; the superscripts ¹²³⁴
mark the same four categories. Because 平 has no letter of its own, the tree adds the superscript from the
Middle Chinese box onward — so kae appears as kae¹ , and you can follow one unbroken tone
category down the whole branch. The Old Chinese box has no marker : it had no tones yet.
Raised ¹ vs full-size 1 調類 vs 調值
The tree writes tone numbers in two different ways , and they are
not the same system . Look at whether the number is raised or full-size.
kae1 Middle Chinese → 1800
Raised, small = tone CATEGORY 調類
Which of the four medieval classes (平上去入) the syllable belongs to. It is a label, not a pitch —
it tells you the historical class and says nothing about how high or low the voice actually went.
It starts at the Middle Chinese box (the first stage with tones) and runs down every historical box,
because the rime books record the category — but the real melody of 1200s speech is not recoverable.
The Old Chinese box has none : tones had not developed yet.
gam1 1900 → 2000 boxes
Full-size, after the syllable = tone VALUE 調值
A modern Jyutping tone number, 1–6 — an actual measurable pitch you can hear and imitate
(1 = high level ˥, 4 = low falling ˨˩ …). See the pitch chart below ↓.
Mandarin's modern boxes use the diacritics ā á ǎ à instead of digits, so full-size tone numbers
in this tree are always Cantonese.
They deliberately don't match. 人 is category ¹ (平) all through the old boxes,
yet its modern Cantonese reading is jan4 . Nothing went wrong: between those stages each of the four
categories split in two by the voicing of the old initial (陰 upper / 陽 lower register), turning
4 categories into Cantonese's 6 tones. 人 had a voiced ny-, so its 平 went to the lower register → tone 4.
One medieval category → two modern tones
平 ¹ → jyutping 1 (陰平, old voiceless) · 4 (陽平, old voiced) — Mandarin 1st / 2nd tone
上 ² → jyutping 2 (陰上) · 5 (陽上) — Mandarin 3rd tone
去 ³ → jyutping 3 (陰去) · 6 (陽去) — Mandarin 4th tone
入 ⁴ → jyutping 1 / 3 / 6 on -p -t -k syllables (the “checked” tones) — in Mandarin, scattered across all four
So a raised ² and a full-size 2 can appear on the very same word and mean different
things: the first says “this was a 上 syllable”, the second says “say it with a mid-rising pitch ˧˥”.
Tone numbers — pitch chart 聲調
Pitch runs high (top of the staff) to low (bottom). The number written after a syllable picks its pitch shape.
Mandarin · 4 tones — pinyin writes them as ā á ǎ à
1 2 3 4
妈 麻 马 骂
mā · high má · rising mǎ · dip mà · fall
Cantonese · 6 tones — jyutping writes them as numbers 1–6
1 2 3 4 5 6
詩 史 試 時 市 是
si1 si2 si3 si4 si5 si6
55 35 33 21 13 22
Numbers under the Cantonese chart are Chao pitch values (5 = highest, 1 = lowest): e.g. 21 starts mid-low and drops. Test sets: 詩史試時市是 = si1–si6 · 妈麻马骂 = mā má mǎ mà.
Old Chinese · Baxter–Sagart 上古漢語
No recordings of c. 1000 BCE survive, so every form is reconstructed — written with a leading * . The brackets and marks say how certain each part is.
* a reconstruction, not a real spelling (天 *l̥ˤin )
ˤ pharyngealisation — a tense, throat-constricted vowel (“Type A”); no ˤ = relaxed (“Type B”). OC’s biggest contrast.
[ ] the bracketed sound is uncertain (天 *l̥ˤi[n] — coda maybe -n or -ŋ)
( ) this sound may or may not have been there
. a reduced pre-syllable before the main one (高 *Cə. kˤaw)
- an affix : prefix 書 *s- ta · suffix 二 *nij- s
C · N an unknown consonant · unknown nasal (something was there)
l̥ n̥ voiceless sonorants (l̥ later hardened to th-, etc.)
q qʰ ɢ uvular stops — further back than k; many became h- or zero
ə · ɨ schwa · high central vowel; ʷ = lip-rounding (火 *qʷʰəjʔ)
↪ Where tones came from: Old Chinese had none. Its endings became the four tone categories — -ʔ → 上 (X), -s → 去 (H), open/nasal → 平, and -p -t -k → 入.
Middle Chinese · Baxter–Sagart 中古漢語
The medieval root of the tree (c. 600 CE, the Qieyun tradition) — one stage younger than Old Chinese above.
ng [ŋ] as in sing
ny [ɲ] palatal n (≈ Spanish ñ)
' [ʔ] glottal stop
h [ɣ] voiced; x = [x] as in Bach
kh th ph aspirated stops (puff of air); sibilants below ↓
j · w palatal (y-) and labial (w-) medial glides
+ [ɨ] high central vowel
ae · ea [æ] low front · retroflex-coloured low vowel
-m -n -ng nasal codas; -p -t -k = stop codas
Sibilants & affricates — exactly how 齒音
These three rows are the same sounds in different mouth positions . The only difference is where the tongue sits ; an h after the letter (tsh, tsyh…) just adds a puff of air, and the voiced ones (dz, dzy…) hum like English z.
Dental — tongue tip at the upper teeth
ts / tsh [ts tsʰ] cats · tsh = same, with a puff of air
dz [dz] voiced — “dds” in adds
s / z [s z] plain s ee / z oo
Palatal — tongue blade flat on the hard palate, tip down (= pinyin j q x)
tsy / tsyh [tɕ tɕʰ] pinyin j / q — “j” in jeep, tongue forward & flat
dzy [dʑ] voiced soft “j”
sy / zy [ɕ ʑ] pinyin x — a forward “sh”; zy = s in “visi on”
Retroflex — tongue curled up & back (= pinyin zh ch sh r)
tsr / tsrh [tʂ tʂʰ] pinyin zh / ch
dzr [dʐ] voiced retroflex “j”
sr / zr [ʂ ʐ] pinyin sh / r
tr / trh / dr [ʈ ʈʰ ɖ] a t /d made with the tongue curled back
New · Mandarin pinyin 漢語拼音
q [tɕʰ] ≈ ch in cheese (tongue forward)
x [ɕ] ≈ a soft sh
j [tɕ] ≈ j in jeep (tongue forward)
zh ch sh retroflex — tongue curled back
z · c [ts] · [tsʰ] as in pizz a / its
r [ʐ] retroflex, between r and zh
ü / yu [y] front rounded (German ü)
e [ɤ] unrounded “uh”
-i after z c s zh ch sh r = a buzzed vowel [ɹ̩]
tones ā á ǎ à — see the pitch chart above ↑
New · Cantonese jyutping 粵拼
j [j] English y (jyut = “yut”)
z · c [ts] · [tsʰ]
gw kw labialised [kʷ kʷʰ]
aa · a long [aː] · short [ɐ]
oe · eo [œː] · [ɵ] rounded mid vowels
yu [y] front rounded
-p -t -k unreleased stop codas (kept from Middle Chinese)
tones 1–6 see the pitch chart above ↑
Sources & method 出處
This is a teaching model , not a research database — and the two layers below have very different reliability .
Sourced Every Middle Chinese form is looked up, character by character, in the Baxter–Sagart table or Wiktionary's Guangyun data — never written from memory. The original 203 core characters were re-audited against those sources (see mc-audit.json): 183 were correct, 10 were corrected, and 9 turned out to have no attested Middle Chinese at all . Characters with no record show “no attested reading” rather than an invented form.
Derived The century-by-century steps are computed by rule from the sourced Middle Chinese: the documented sound laws (devoicing, entering-tone loss, -m → -n, palatalisation, the Cantonese kʷ → f shift…) applied in their conventional order, anchored at both ends by the Middle Chinese form and the attested modern reading. Reproducible — but still model output, not attested : no source records a per-character pronunciation for a given century. The oldest core words keep their earlier hand-curated stages.
HSK 3 · northern branch Not interpolated at all. For the 154 characters used only by HSK 3 words, the Mandarin stages are produced by running nk2028's published derivation schemas (tshet-uinh + tshet-uinh-examples, the library behind the Tshet-uinh Deriver ) over each character's Guangyun 廣韻 position. 北宋 Northern Song sits on the 1200 tick and 《中原音韻》 (1324) on the 1300 tick — each on the nearest tick at or after its source's date — and the 1324 form is then held, not animated , through 1400–1700, because no source in this set attests a change there. Hover any of those boxes for its 音韻地位 and its ʼPhags-pa spelling in the 《蒙古字韻》 (1269) , which, being alphabetic, records sounds rather than rime categories. The reading is chosen by matching the schema's own Baxter output to the Middle Chinese in the root box, so root and timeline always describe one syllable.
Old Chinese Baxter & Sagart, Old Chinese: A New Reconstruction (Oxford Univ. Press, 2014)
Middle Ch. Baxter, A Handbook of Old Chinese Phonology (1992) — the transcription; based on the Qieyun 切韻 (601 CE) rime tradition
Late MC / Mand. Pulleyblank, Lexicon of Reconstructed Pronunciation… (UBC Press, 1991)
Old Mandarin the Zhongyuan Yinyun 中原音韻 (Zhōu Déqīng, 1324)
Cantonese Bauer & Benedict, Modern Cantonese Phonology (1997)
Overview Norman, Chinese (Cambridge Univ. Press, 1988)
Sound changes the 26 named changes behind the click-a-box explainer follow Historical Chinese phonology (Old → Early MC → Late MC → Mandarin), Middle Chinese , Cantonese phonology , Checked tone and Four tones of Middle Chinese — archived locally in data/phonology/
⏳ Timeline the 14 entries draw on the same archived articles and the works they cite: Baxter (1992) for labiodentalisation, Bauer & Benedict (1997) and Norman (2003) for the length-conditioned split of the upper entering tone and the Guangzhou/Hong Kong tone counts, Bauer & Benedict (1997) and Yip & Matthews (2001) for the n ~ l merger, Baker & Ho (2006) for kʷ- → k-, and 何大安 (1994) 《「濁上歸去」與現代方言》. The dated loss of the Cantonese alveolar ~ alveolo-palatal contrast rests on the dictionaries that recorded it while it lived: Williams (1856), Cowles (1914), Meyer & Wempe (1947) and Chao (1947)
Readings Pinyin: Xiàndài Hànyǔ Cídiǎn 现代汉语词典 · Jyutping: LSHK 粵拼 scheme (1993)
Divergent words all 48 pairs checked against the Unihan Database (jyutping/pinyin), CC-Canto and CC-CEDICT (glosses & readings), plus Baxter–Sagart for their Middle/Old Chinese — see cantonese-audit.json. For the 19 everyday pairs added last, the Middle Chinese and the 廣韻 glosses quoted in the notes (企 企望也 “stand on tiptoe and look”, 站 俗言獨立 “colloquially, to stand alone”, 匙 匕也 “a spoon”…) come from nk2028's Guangyun data . Old Chinese is shown only for single-character words: the multi-character entries are post-classical compounds, so a compound-level Old Chinese root would be a fiction
Word lists HSK 1–4 vocabulary from hsk.academy , archived locally as data/hsk1–4.tsv; glosses condensed from CC-CEDICT , jyutping from CC-Canto and Unihan . Parts of speech are assigned by rule and then hand-corrected (data/hsk3-overrides.tsv)
廣韻 readings for the HSK 3 characters outside the Baxter–Sagart table, the Middle Chinese was taken from Wiktionary's per-character Qieyun /Guangyun modules (Module:zh/data/ltc-pron/字, giving initial · rime · division · tone · fǎnqiè) and converted to Baxter transcription with a local port of Module:ltc-pron/baxter — modules and conversion archived in data/ltc/. Nine characters (啊 搬 蛋 筷 辆 啤 卡 爷 嘴) have no entry in that dataset and are shown with no Middle Chinese rather than a guess
Derivation engine nk2028 , Tshet-uinh Deriver (nk2028.shn.hk/tshet-uinh-deriver) and its libraries tshet-uinh 0.15.4 / tshet-uinh-examples 20260629 / tshet-uinh-deriver-tools 0.2.0, archived in data/nk2028/. Its Zhongyuan Yinyun schema follows 寧繼福 《中原音韻表稿》 (1985) and 薛鳳生 《中原音韻音位系統》 (1990). Note that even these stages are systematic derivations from rime-book categories, not recordings — a large upgrade on interpolation, but the ±a-century caveat stands
Compiled and interpolated by hand from these sources for illustration — treat the intermediate forms as informed approximations, not citable data points.