Key takeaways
- Jyutping numbers six contrasting tones from 1 to 6, and the structure is easier to hold than the list: three level tones at three heights, two rising tones, and one low falling tone.
- The 'nine tones' figure counts syllables ending in -p, -t or -k separately; those checked syllables reuse the pitch levels of tones 1, 3 and 6 rather than adding new ones.
- Tone is relative to a speaker's own pitch range, not an absolute frequency, so perfect pitch is irrelevant and contour matters more than height.
- The contrasts that cause most beginner errors are 2 against 5, 3 against 6, and 4 against 6 — pairs that differ in height or starting point rather than in shape.
- Tone is part of the word, so it has to be learned at first exposure; correcting a tone months later is far harder than storing it correctly the first time.
Most explanations of Cantonese tones hand you a numbered list of six pitch descriptions and leave you to it. The list is accurate and almost useless, because it presents six independent facts to memorise when what is actually there is one small structure with six positions in it.
Learn the structure and the list stops being a list.
What tone is doing
In English, pitch belongs to the sentence. You raise your voice at the end of a question, you drop it at the end of a statement, and you can say the word "cat" on any pitch you like without it becoming a different animal.
In Cantonese, pitch belongs to the word. The same string of consonants and vowels said on a different pitch pattern is a different word — not an emphasised version of the same one, not a regional accent, a different entry in the dictionary. Tone is a property of the word in exactly the way the vowel is.
That has one immediate consequence for how you study: you cannot learn vocabulary now and add tones later. A word stored without its tone is stored wrong, and the correction is harder than the original learning would have been.
The six tones, as a structure
Jyutping numbers the tones 1 to 6 and writes the number after the syllable: si1, si2, si3. Here is the usual description.
| Jyutping | Usual description |
|---|---|
| 1 | high level |
| 2 | rising |
| 3 | mid level |
| 4 | low falling |
| 5 | low rising |
| 6 | low level |
Now the part that makes it memorable. Those six are not six unrelated shapes. They are:
- Three level tones at three heights — 1 high, 3 mid, 6 low. Same flat shape, stacked.
- Two rising tones — 2 and 5. Same rising shape, starting from different heights.
- One low falling tone — 4, the lowest thing in the system, often described as creaky in ordinary speech.
Six tones, three shapes. Once you hear the system that way, "which tone was that?" becomes two smaller questions — what shape and how high — and both are easier to answer than the original.
A caution that belongs here rather than in a footnote: descriptions differ between references. Some transcribe tone 1 as high falling for certain speakers or in certain positions, and exact pitch heights vary by region, by individual and by how carefully someone is speaking. Some linguistic descriptions use Chao's five-point pitch scale to write these values numerically. Treat any tone table, including this one, as a map rather than a measurement.
Tone is relative, which is why perfect pitch is beside the point
The single most discouraging myth about tonal languages is that you need a musical ear.
Cantonese tone is defined relative to the speaker's own range. A child, a woman with a high voice and a man with a low one all produce recognisable tone 1 at completely different absolute frequencies. What identifies the tone is where it sits within that speaker's range and what shape it traces.
So you are not learning notes. You are learning a small set of patterns that get rescaled to whoever is talking — which is a normal perceptual skill, not a talent.
The classic demonstration
The standard teaching example is the syllable si, which supports all six tones with six unrelated meanings:
| Jyutping | Character | Meaning |
|---|---|---|
| si1 | 詩 | poem |
| si2 | 史 | history |
| si3 | 試 | to try, to test |
| si4 | 時 | time |
| si5 | 市 | market, city |
| si6 | 是 | to be |
One note on the last row: 是 is the formal and written word. In spoken Cantonese, "to be" is 係 (hai6) — same tone, different word — and you will meet 是 mostly when reading Standard Written Chinese aloud. That distinction between spoken and written registers is one of the things beginners meet earlier than they expect, and it is covered in our beginner's guide to Cantonese.
The set is not a party trick. It is the proof that tone is lexical: nothing else about these six syllables differs.
Why some sources say nine
You will see Cantonese described as having nine tones, and the disagreement is not about the facts.
Cantonese syllables can end in -p, -t or -k. Those endings stop the syllable short — the vowel is cut off rather than allowed to run. An older analytical tradition, inherited from classical Chinese phonology, counts these "entering tone" syllables as their own category, and since they occur at three pitch levels, that adds three to the count. Six plus three is nine.
Modern descriptions do not add them, because the three pitch levels involved are the ones already used by tones 1, 3 and 6. Nothing new is being distinguished. Jyutping therefore writes checked syllables with the same numbers:
- 一 (jat1) — "one", checked, at the tone 1 level
- 百 (baak3) — "hundred", checked, at the tone 3 level
- 十 (sap6) — "ten", checked, at the tone 6 level
Six and nine are two ways of counting one system. If a course teaches you nine, you are not being misled; you are being taught with older terminology.
Checked syllables are worth practising deliberately for a different reason: they are short. There is less time for the pitch to be audible, so the contrast has to be caught quickly. A useful minimal pair here is 識 (sik1, "to know") against 食 (sik6, "to eat") — identical apart from tone, both cut off by the same final consonant, and both extremely common.
The contrasts that actually cause errors
Not all fifteen possible pairs are equally hard. Practice time is better spent on the three that generate most beginner mistakes.
Tone 2 against tone 5. Both rise. The difference is where the rise starts and finishes, not the shape. This is comfortably the hardest pair for most learners, and it matters because tone 5 carries a large share of the most common words in the language — 我 (ngo5, "I"), 你 (nei5, "you"), 佢 (keoi5, "he/she") are all tone 5.
Tone 3 against tone 6. Both level. The difference is purely height, with no shape cue to fall back on. Height differences are harder to judge than shape differences, especially across speakers.
Tone 4 against tone 6. Both low. Tone 4 falls and tone 6 stays flat, but at the bottom of a speaker's range the difference is small and often blurred by creak.
By contrast, tone 1 is usually the easiest tone to identify, because it sits clearly above everything else in the system. If you can reliably hear tone 1 versus everything, you have already solved a fifth of the problem.
The practical instruction is unglamorous: drill the confusable pairs, not the whole set. Six-way identification exercises spread your attention evenly across contrasts you have already mastered and contrasts you have not.
One complication worth knowing about early
Cantonese has a phenomenon usually called changed tone, where a syllable in certain words and certain contexts is pronounced with a different pitch from its dictionary tone — commonly a rising one. It turns up in familiar or derived vocabulary, in some compounds, and in some ways of naming people and things.
There is no simple rule you can apply on sight, and it is not a mistake or sloppiness. The honest advice for a beginner is: learn that it exists so you are not confused when a word you know comes out differently, and learn the individual cases as vocabulary rather than trying to derive them. A course that flags these words as they arrive saves you the experience of concluding your ear is broken.
How to practise tone without a teacher in the room
Five things that work, in the order they pay off.
Learn tone with the word, always. Never write down a Cantonese word without its Jyutping number. A vocabulary list of bare syllables is a list of words you will have to relearn.
Drill in pairs, not in sets of six. Take one confusable pair — 2 and 5, say — and work only on that until identification is fast. Then take the next pair. Six-way drills feel more advanced and teach less.
Separate perception from production. These are two skills and they improve at different rates. Being able to hear that two recordings differ does not mean you can produce the difference, and being told your production is wrong does not tell you what to change. Work on identification first; it is the one you can verify alone.
Record yourself and leave a gap. Listening back immediately, you hear what you intended. Listening back the next day, you hear what you said. That delay is doing real work.
Use the words in sentences early. Tone in isolation is a laboratory exercise. Tone at conversational speed, inside a phrase, with your attention on meaning rather than pitch, is the actual skill — and it is noticeably harder than the isolated version, which is why doing only isolated drills produces a plateau.
Learn Cantonese for Beginners is sequenced around the first two of those. Initials, finals and syllables come first, then tone contrasts, before conversation starts, so pronunciation is a foundation rather than a later correction. Where your device provides a Cantonese voice the app uses it and falls back clearly when it does not. Review modes — Daily Review, Speed Run Review, Mistake Rescue, Listening Blitz and Dictation Mode among them — pull from material you have already covered, and a missed-questions list collects what you keep getting wrong, which is exactly where a confusable tone pair ends up.
Two limits stated plainly. No app can hear you the way a fluent speaker can, so it cannot tell you which of six tones you actually produced. And recognising a contrast in a quiet room is easier than producing one mid-sentence, which is easier than either in a noisy restaurant. This is foundation work, not a substitute for a teacher or a conversation partner.
Is this the hardest part of Cantonese?
It is the part that frightens people most, which is not the same thing.
The US Foreign Service Institute places Cantonese in Category IV, its hardest group, at 88 weeks of full-time instruction (2,200 class hours) for its own students — a figure that describes intensive classroom training for diplomats, not app study, and that covers the whole language rather than its tone system. Tone is one component of that, and it is a component with an unusual property: it front-loads.
Tones are hardest in weeks one to eight, when every word is new and nothing is automatic. They get quietly easier after that, not because your ear improves dramatically but because context starts doing more of the work and because the words you use most often become fixed. The learners who struggle longest are usually the ones who postponed tone in month one and are now carrying a few hundred words that need repairing.
Which is the entire argument for doing this first. Tones are not the hard part of Cantonese if you learn them with the vocabulary. They become the hard part if you do not.
If you are weighing Cantonese against other tonal languages before committing, the Thai script takes a completely different approach — five tones that are calculated from spelling rather than memorised per word — and the rest of our language learning apps include options with no tones at all.
Good to know
Frequently asked questions
Does Cantonese have six tones or nine?
Do I need perfect pitch to learn Cantonese?
Which Cantonese tones are hardest to tell apart?
What happens if I say the wrong tone?
Can an app teach me tones?
Does Learn Cantonese for Beginners cover tones before conversation?
Sources
- Jyutping FAQ (opens in a new tab)The Linguistic Society of Hong Kong — accessed
- Unicode Character Code Charts (opens in a new tab)The Unicode Consortium — accessed
- Foreign Language Training (opens in a new tab)U.S. Department of State, Foreign Service Institute — accessed