A tone is not a melody. Each of the six is a bundle of three things at once, and pitch is only the first of them: the contour, the voice quality the throat is using while it happens, and how long the syllable lasts. The two marks nobody can separate at the start, the hook and the tilde, are close in pitch and far apart in the other two.
This is why humming the tones works and singing them back does not. A hummed contour drops the creak, and the creak is half of what a listener is using.
The six, and the one word that carries all of them
- ma, mà, má, mả, mã, mạ, ghost, but, mother, tomb, horse, rice seedling. One syllable, six tones, six unrelated words, and this is the standard demonstration because it is genuinely how the language works.
- ngang, huyền, sắc, hỏi, ngã, nặng, the names of the six: level, falling, rising, dipping, broken and heavy.
- the tone marks are printed in all ordinary text, unlike the vowel marks of Arabic or Hebrew, so nothing has to be guessed.
- nặng is a dot underneath, the only mark below the letter, and it shortens the syllable sharply as well as lowering it.
- the south merges hỏi and ngã into a single tone, which is why southern speakers make spelling errors between them and why the southern system has five.
- the tone is part of the word. Saying the syllable with the wrong tone is not an accent, it is a different word, and this is the adjustment that takes longest.
What the six are actually doing
The unmarked one sits level in the middle of the range and does nothing, which makes it the hardest to hold when the sentence around it is moving. The grave mark falls slowly from mid to low and stays breathy. The acute rises sharply to the top of the range. The hook dips and comes back up, smoothly, with no break in it. The tilde also rises, but it snaps in the middle, and that break is a closure in the throat rather than a change of note. The dot below is low, short and squeezed off at the end.
Read that list again and notice how much of it is not pitch. Two of the six are defined by what the larynx does, one by length, and only three by the shape of the line alone.
The restriction that halves the work
A syllable ending in p, t, c or ch can carry only two of the six: the acute or the dot below. Nothing else is possible, because those endings stop the airflow and there is no room left for a contour to develop. So for a large slice of the vocabulary the choice is between two tones rather than six, and a learner who knows this stops trying to hear four tones that were never candidates.
That is the sort of fact worth finding early in any language, because it converts an apparently large system into a smaller one. Nothing else in Vietnamese repays a week of study as well.
You heard it and cannot name it
The first question is not which note it was
Length and breaks are audible before a contour has finished, so the fork that comes first is the one about the end of the syllable.
- It stopped shortThe dot below, or a checked syllableLow, squeezed off in the throat, over before a contour had room to develop
- If it ends in p, t, c or ch, only two of the six were ever possible
- Otherwise it is the dot below
- Length is doing the identifying here, not pitch
- It ran its full lengthNow listen for a break, and only then for the shapeFour candidates remain, and one cue separates two of them from the other two
- A break in the middle: the tilde
- A smooth dip and return: the hook
- Flat and mid, with no movement: unmarked
- Falling and breathy: the grave. Rising to the top: the acute
Nobody runs this consciously at speed. It is a way of listening to recordings until the cues attach themselves to the sounds, after which the question stops being asked.
Where it stalls: tones that survive the card and die in the sentence
Around the second or third month a learner appears who can produce all six on an isolated syllable, accurately, on demand, and loses every one of them the moment the syllable is inside a phrase. They usually conclude that they need more tone drilling, and more tone drilling on isolated syllables makes it worse.
The mechanism is that a tone in running speech is not the same object as a tone in isolation. Its contour is compressed into less time, its starting height is pulled towards whatever the previous tone ended on, and the pitch of the phrase as a whole is riding on top of it. A tone drilled alone is a deliberate pitch gesture the throat performs with its full attention. In a sentence there is no full attention available, and the gesture has to be smaller and automatic.
The exit is to make the smallest unit of practice a real phrase with a fixed frame around it, so the tone is produced under the same conditions it will be used in. Say the target word inside the same short carrier sentence twenty times, then swap the word for its minimal pair and hear what changed while everything else stayed still.
Two things worth using
Praat, the acoustic analysis program from the University of Amsterdam that phonetics departments have used since the early 1990s and still give away, draws your pitch contour under the waveform. Record a native speaker's syllable, record your own, and look at the two traces. It replaced the teacher saying that was not quite right and trying again, which tells you that you failed without telling you in which direction, and it costs nothing.
The second is a carrier sentence you write once and never change, with a single slot in it. It replaced the isolated syllable flashcard, which trains a tone that only exists alone, and it is free to make and has to be made by you, because the frame has to be a sentence you would actually say.
The mistake: learning the tone as a name
The marks have names, the names get memorised, and then recognition runs through the name: hear the syllable, identify the mark, recall which tone that is. That chain is far too slow for speech and it breaks under any pressure at all. The tone has to be stored as part of the sound of the word, so that the wrong tone sounds wrong the way a wrong vowel does in your own language.
The practical test is whether you can hear an error without being able to say what it was. If a friend says a word with the wrong tone and you flinch before you can name the mark, the storage is right. If you have to think, it is still notation. Card design decides which of the two you end up with, which is making cards that teach rather than cards that pile up, and the material you meet the tones in has to be simple enough that attention is free for them, which is material just past your level, and why that is the sweet spot. An hour with a speaker who will correct one thing repeatedly is worth more here than in most languages, which is what a teacher is worth, and what you can do alone.
The next question is what the tone is sitting on. A Vietnamese syllable has a fixed shape with a small set of possible endings, and those endings decide which tones can appear at all, in final consonants that are held rather than released. And the six are six only in the north, in tones and vocabulary that differ by where you are.