The Four Ancestors
Around the 7th century CE — the era of the Qieyun 切韻 rime dictionary (compiled in 601 under the Sui) and of the Tang dynasty that made it the standard — Chinese had four tone categories: 平 (píng, level), 上 (shǎng, rising), 去 (qù, departing), and 入 (rù, entering). The entering tone (入声) was not a pitch shape but a syllable structure category: all its syllables ended in a short stop consonant (-p, -t, or -k). These four categories held all the syllables of Middle Chinese.
The Voicing Split — One Rule to Rule Them All
Middle Chinese had a rich system of initial consonants, many of which came in voiced/voiceless pairs: p- vs b-, t- vs d-, k- vs g-, etc. Over the early medieval period, voiced initials disappeared or merged differently in each dialect, but not before leaving a lasting imprint: syllables with originally voiced initials developed lower pitch, while those with voiceless initials developed higher pitch. This is the Yīn-Yáng split (阴阳分化).
The result is mechanical but spectacular: each of the original four tones split into a Yin (阴/upper) register and a Yang (阳/lower) register, simply based on what kind of consonant the syllable used to start with. Four tones × two registers = eight potential tone slots. Dialects then merged some slots, split others differently, and the final count diverges wildly across the Sinitic family.
The Cantonese Case: 9 Tones (or 6+3)
Cantonese is the canonical showcase. The six live tones (陰平, 陽平, 陰上, 陽上, 陰去, 陽去) cover the old 平/上/去 categories after the Yin-Yang split — 6 contour tones with distinct pitch shapes. But Cantonese also preserved the entering tone (入声). The Yin-Yang split gave it two checked tones, and a later Yue-internal split of the upper one — conditioned by vowel length, not by the old initial — added a third, yielding 3 checked tones (陰入, 中入, 陽入) with the same pitch heights as the upper, mid, and lower live tones but cut short by a final stop. The traditional count is therefore 9 tones, but no romanisation actually writes nine. Yale distinguishes only six, spelling a checked syllable exactly like the high-flat, mid-flat or low-flat tone whose pitch it shares; Jyutping, the system used on this map, also numbers 1–6 only (一 jat1, 目 muk6). The digits 7, 8 and 9 for 陰入/中入/陽入 live on in traditional tone charts and in the Cantonese Pinyin (教院式) system — bookkeeping labels for the history rather than three extra pitches.
Mandarin: 4 Tones, the Entering Tone Vanished
Standard Mandarin simplified dramatically. The voiced initials disappeared early and thoroughly; the Yin-Yang split did occur but several registers merged. The entering tone (入声) lost its final stops and its syllables were redistributed across the four remaining tones — almost randomly, creating the notorious unpredictability of which modern Mandarin syllable was originally a 入声 character. The result: a clean 4-tone system (阴平, 阳平, 上声, 去声), the sì shēng 四声 that every Chinese learner memorises.
Min Nan: 7 or 8 Tones, Entering Tone Alive and Doubled
Hokkien/Taiwanese (闽南语, Min Nan) preserved the entering tone with its final stops intact (-p, -t, -k). Like Cantonese, it applied the Yin-Yang split to the entering tone, giving 2 entering tones (陰入 and 陽入). The live tones underwent further mergers and splits, yielding a base of 7 tones in Standard Southern Min. Min Nan is also famous for its sandhi system: in natural speech, the tone of a syllable changes depending on its position in a phrase, so learners must memorise both the citation tone and the sandhi tone for every word.
Hakka, Wu, and the Full Spectrum
Hakka (客家话) dialects typically have 5–7 tones, preserving the entering tone in some varieties. Wu (吴语, including Shanghainese) kept voiced initials longer than any other major Sinitic branch, meaning the Yin-Yang split is still partially phonetically audible. Some Wu dialects count 7–8 tones. Linguists sometimes describe Wu as offering a living snapshot of what Middle Chinese voicing contrast sounded like.
Cantonese is unusually retentive of Middle Chinese finals and tone categories, but its nine tones are not simply the Tang four doubled. Eight come from the Yin-Yang split; the ninth is a later, Yue-internal development, in which vowel length split the upper entering tone (陰入) into 陰入 and 中入.