Mandarin and Cantonese Tones Ask Your Ear Different Questions
Mandarin has four tones and Cantonese has six, and the count is the least interesting difference between them. Mandarin separates its tones mostly by direction, with one level tone and three that move. Cantonese has three level tones that differ only in height. An ear trained on one language is being asked a different question by the other, and the habit that makes you good at Mandarin is the habit that makes Cantonese level tones invisible.
The count is the least of it
Cantonese is often said to have nine tones. Modern descriptions of the Hong Kong variety give it six, and both numbers are defensible because they are counting different things. Six is the number of tones that can distinguish two otherwise identical syllables. The extra three in the traditional count are the checked tones, which occur only on syllables ending in a p, t or k, and which are short realisations of three tones the language already has rather than three further contrasts.
So the honest comparison is four against six. That gap sounds like the whole story and it is not. Adding two categories to a system is a matter of more practice. Changing what the categories are made of is a different kind of work, and that is what happens when you move between these two languages.
The two systems side by side
The numbers below are Chao pitch values, the five-point notation both traditions use, where 5 is the top of a speaker's comfortable range and 1 the bottom. Sources vary by a point here and there, particularly on the Cantonese rising tones, and the shapes are what matter.
| Language | Tone | Pitch value | Shape | Example |
|---|---|---|---|---|
| Mandarin | 1 | 55 | High level | 妈 mā |
| Mandarin | 2 | 35 | Rising | 麻 má |
| Mandarin | 3 | 214 | Dipping, low through most of it | 马 mǎ |
| Mandarin | 4 | 51 | Falling, steeply | 骂 mà |
| Cantonese | 1 | 55 | High level | 詩 si1 |
| Cantonese | 2 | 25 | High rising | 史 si2 |
| Cantonese | 3 | 33 | Mid level | 試 si3 |
| Cantonese | 4 | 21 | Low falling | 時 si4 |
| Cantonese | 5 | 23 | Low rising | 市 si5 |
| Cantonese | 6 | 22 | Low level | 是 si6 |
Read the shape column on its own. Mandarin has one level tone and three that move, and the three that move go in visibly different directions. Cantonese has three level tones, two risings that differ mainly in where they start, and one fall. Half the Cantonese system is answering the question how high and almost none of the Mandarin system is.
Tingyin is a Mandarin tone-listening trainer and covers only the top half of that table: a clip plays, you pick one of the four tones, it tells you immediately. Every one of its 637 clips is a human recording carrying its source, speaker, licence and a sha256, 622 under CC BY-SA 4.0 and 15 under CC0, and levels 1 and 2 are free with level 1 needing no account. It has no Cantonese material and is not the tool for the bottom half.
Three level tones is the real gap
Cantonese tones 1, 3 and 6 are all flat. Nothing about their movement distinguishes them, because none of them moves. The only thing separating 詩 si1 from 試 si3 from 是 si6 is where the flat line sits inside the speaker's own range, and that range differs between a child and an adult by more than the whole distance between the three tones.
This is precisely the judgement Mandarin trains you to distrust. A Mandarin learner is told, correctly, that absolute pitch is not the cue and that a first tone from a low voice can sit below a fourth tone from a high one. The reliable question in Mandarin is which way the pitch went. Carry that heuristic into Cantonese and three of its six tones collapse into one undifferentiated flat thing.
What Cantonese requires instead is a two-step judgement: first establish where this particular speaker's range sits, from whatever has already been said, then place the flat syllable inside it. Native listeners do this in a syllable or two and it feels like nothing. For a Mandarin-trained ear it is a new operation, not a harder version of an old one.
There is a live complication worth knowing about. A well-documented change is under way in Hong Kong Cantonese in which some younger speakers merge tone pairs, most often 2 with 5 and 3 with 6. That means a learner may be receiving inconsistent evidence about exactly the contrasts that are hardest to build, depending on who they listen to. It also means a six-way description is a description of the system, not a promise about any given speaker.
Checked syllables: a tone with no time to happen
Cantonese keeps the final stop consonants that Mandarin lost. A syllable can end in p, t or k, the stop is unreleased, and the whole syllable is short: 一 jat1, 八 baat3, 六 luk6. Those are the checked tones of the traditional nine-tone count, and each one is a shortened version of tone 1, 3 or 6.
For a Mandarin listener this is genuinely unfamiliar territory, and not because the tones are new. It is because there is so little time. A Mandarin syllable that ends in a vowel or a nasal gives the pitch a long window to trace a shape in; a checked syllable gives it a fraction of that and then stops dead. The tone can only be level, because nothing else fits.
The Mandarin equivalent of that pressure does exist, and it is instructive. Mandarin tones do not flatten in fast speech the way people expect, because the pitch excursion holds even as the syllable shortens. A checked Cantonese syllable is the opposite arrangement: the syllable is short by design and the tone system has adapted by using only shapes that fit.
What actually transfers
Someone who already speaks one of these languages has a real head start on the other, and it is not in the tone shapes. It is in three habits that a speaker of a non-tonal language has to build from nothing.
- Pitch is part of the word. This is the conceptual hurdle that takes English speakers the longest. Someone who grew up with either language never has to be persuaded of it, and never has to unlearn the reflex of hearing a pitch movement as an emotion or an emphasis.
- Speakers get normalised automatically. Both languages require reading pitch relative to the person talking rather than in absolute terms. That machinery is already built and it carries over whole.
- One close pair will be the problem. Both systems have a contrast that stays difficult long after the others are solved. In Mandarin it is the second tone against the third. In Cantonese it is usually 2 against 5, or 3 against 6. Knowing that a system has one stubborn pair, and that the fix is drilling the pair against itself rather than more mixed practice, is experience that transfers directly.
What interferes, in each direction
Interference here comes from categories that look like matches and are not. The two directions have different failure modes, and it is worth naming them separately.
Cantonese speaker learning Mandarin. Tone 1 lines up almost exactly and tone 2 is close enough to borrow. The fourth tone is where it goes wrong: a Mandarin fall runs from the top of the range to the bottom, and the nearest Cantonese shape, tone 4, is a shallow fall in the low part of the range. Borrowing it produces a Mandarin fourth tone that starts too low and does not travel far enough, which native Mandarin listeners hear as hesitant rather than as wrong. The third tone has no Cantonese counterpart at all as a dipping contour and usually gets mapped onto tone 5 or tone 4.
The other Mandarin-specific hurdle is alternation. Mandarin changes its tones according to what follows: two third tones in a row and the first becomes a rise, and 一 and 不 shift depending on the next syllable. Cantonese has changed tone as well, but it is lexical and morphological rather than an automatic rule applied on the fly, so Mandarin sandhi is a genuinely new category of thing to learn rather than a variant of something familiar.
Mandarin speaker learning Cantonese. The level triple is the work, and the reflex to fight is reaching for direction when there is none. A second habit to break is the assumption that a syllable is long enough to hear a shape in, which checked syllables violate. And there is a trap in the character correspondences: the historical relationship between the two tone systems is regular, since both descend from the same Middle Chinese categories, but it is a relationship between historical classes rather than a table you can run in real time. Learners who try to convert a Mandarin tone into a Cantonese one in their head are usually right often enough to keep doing it and wrong often enough for it to hurt.
Learning Mandarin where Cantonese is spoken
Anyone learning Mandarin in Hong Kong or Guangdong has an input problem that learners elsewhere do not. Much of the Mandarin you hear day to day is spoken by people whose first language is Cantonese, and it carries systematic transfer: a flatter fourth tone, a third tone realised as a plain low tone, and retroflex initials produced further forward. None of that makes it bad Mandarin. It makes it a variety, and a variety is a poor thing to build your first perceptual categories from, because your categories will end up shaped around it.
The practical answer is to make sure some of your listening comes from a source you have chosen rather than from whoever happens to be nearby. That is a question about provenance, and provenance has limits worth knowing: Tingyin's manifest records source, speaker, licence and hash for every clip, and it does not record which variety of Mandarin each speaker grew up with, which is the gap set out in what a manifest can and cannot tell you about an accent.
For the reverse trip, a Mandarin speaker working on Cantonese, the tooling is thinner but it exists. Pleco shows Cantonese readings in Jyutping alongside its Mandarin entries, and words.hk is a free Cantonese dictionary with recorded audio, built by Hong Kong speakers. Neither is a perception drill. Building the deck yourself in Anki, three level tones on one syllable and nothing else varying, gets you the forced-choice exercise that actually moves a category.
Frequently asked questions
What is the best way to learn Cantonese tones if I already know Mandarin?
Start with the three level tones and ignore everything else until they separate, because they are the part your Mandarin training actively works against. Drill them on one syllable at a time, from a single speaker, so that the only variable is height inside that person's range. words.hk gives you free recorded audio from Hong Kong speakers to build such a set from, and Anki will hold it if you are willing to assemble the deck. Mandarin resources will not help here, and Tingyin covers only the Mandarin four.
How do I hear the difference between the Cantonese level tones?
By anchoring the speaker before you judge the syllable, rather than judging the syllable on its own. Tones 1, 3 and 6 are all flat and differ only in height, so a flat syllable heard in isolation genuinely does not carry enough information; what tells you is where it sits relative to pitches the same voice has already produced. This is the opposite of the Mandarin instruction to ignore absolute pitch and attend to direction, which is why transferring that instruction is the single commonest mistake in this direction.
What is the difference between saying Cantonese has six tones and nine tones?
Six counts the tones that can distinguish two otherwise identical syllables, and nine counts the traditional categories, which split three of those six according to whether the syllable ends in a stop consonant. A checked syllable such as 一 jat1 or 六 luk6 ends in an unreleased p, t or k and is much shorter, so its tone is a compressed version of tone 1, 3 or 6 rather than a seventh, eighth or ninth contrast. Both counts describe the same language, and the six-way count is the one that tells you how many categories you have to learn.
Is it safe to learn Mandarin and Cantonese at the same time?
It is safe in the sense that nothing gets permanently damaged, and it is slower than doing one and then the other, because the two systems interfere in exactly the places that are hardest anyway. The specific risk is that near-matching categories get merged: Mandarin tone 1 and Cantonese tone 1 are close enough to borrow, which encourages borrowing the fourth tone from Cantonese tone 4, where the shapes are not close at all. If you do run both, keep the perception work separate by session and by speaker, so each language gets its own set of anchors.
Do Cantonese speakers find Mandarin tones easy?
Easier than English speakers do, and not easy. The largest hurdle for a learner from a non-tonal language, accepting that pitch is part of the word rather than a mood, is already cleared, along with the habit of reading pitch relative to the person talking. What remains is real work: the Mandarin fourth tone falls further than anything in Cantonese, the dipping third tone has no counterpart, and the automatic sandhi rules are a kind of alternation Cantonese does not have.