Every foreigner learning Vietnamese is shown the same chart sooner or later: six tones drawn as lines on a pitch scale. Ma (ghost) is level, mà (but) falls, má (mother) rises, mả (grave) dips, mã (horse) rises with a break, mạ (rice seedling) drops short. It is a useful chart, and it is also slightly misleading, because it suggests that a Vietnamese tone is just a melody. For half the country it isn’t.
I grew up hearing the Hà Nội tones as “obvious”, and only started to understand them when I read what phoneticians actually measure. Three things surprised me: two of the tones are partly made with creaky voice rather than pitch, the North and South use different cues to tell tones apart, and the whole six-tone system is the fossil of consonants Vietnamese lost long ago.
Two of the six tones live in the throat
In the standard description of Hà Nội Vietnamese, James Kirby’s 2011 Illustration of the IPA in the Journal of the International Phonetic Association, tone and voice quality are fused. Ngã is a rising tone with a glottal interruption or stretch of creaky voice in the middle; nặng is a low, short, falling tone that ends in creak or a glottal closure. Even hỏi, the dipping tone, tends to be more constricted than the plain tones. Phonetic work on Hà Nội speech has documented several kinds of creak in these tones: vocal fry, laryngealisation, even period doubling.
That matches what I hear when I listen carefully to myself. Say mã slowly in the northern way and there is a catch in the middle, almost as if the syllable were cut in two. Say mạ and the voice drops and closes off. If you produce the right pitch shape with a smooth, clear voice, a northerner will often still hear something slightly off.
North listens to the throat, South listens to the melody
Marc Brunelle’s 2009 study in the Journal of Phonetics compared how Northern and Southern listeners perceive tones. The finding, repeated in much of the later literature: Northern listeners use a combination of pitch and voice quality, while Southern listeners rely essentially on pitch. Southern speakers can imitate the northern creak, but everyday Southern speech doesn’t use voice quality to keep words apart.
The South also has one tone fewer in practice. Hỏi and ngã have merged into a single dipping-rising tone, so sửa (to fix) and sữa (milk) sound the same in Sài Gòn, and spoken Southern Vietnamese works with five tones. The merger shows up in spelling: hỏi-versus-ngã mistakes are a classic headache for southern schoolchildren, because the ear gives them no clue. A 2022 Yale thesis by Jewell tested this and found that speakers who merge the two tones still identify them above chance when a non-merging speaker says them. People can hear a distinction they no longer make.
Where the tones came from
Proto-Vietic, the ancestor of Vietnamese and Mường, had no tones. The classic explanation is André-Georges Haudricourt’s 1954 paper De l’origine des tons en vietnamien, which compared Vietnamese with its Mon-Khmer relatives and showed that the tones grew out of consonants:
- Syllables that ended in a glottal stop developed the tones now written sắc and nặng.
- Syllables that ended in -h (itself often from older -s) developed hỏi and ngã.
- Syllables with neither became ngang and huyền.
That gave three tone categories. Then each split in two depending on whether the initial consonant was voiceless or voiced, since voiced consonants lower the pitch of the vowel after them. Three times two is six. The voiceless “high” series is ngang, sắc, hỏi; the voiced “low” series is huyền, nặng, ngã.
Later scholars have refined the story. Graham Thurgood argued in 2002 that the model should be framed in terms of laryngeal voice quality rather than consonant segments alone, which fits nicely with the creak that is still audible in ngã and nặng today. But the core idea, that Vietnamese tones are compressed history, is widely accepted. One odd consequence: ngã now sounds high in Hà Nội, yet by origin it belongs to the low series.
Why the old grid still matters
This history isn’t just trivia. The high/low split, which Vietnamese schools teach as bổng and trầm, still governs a large class of native words: reduplicatives (từ láy), where a syllable is echoed with a change. The two halves usually take tones from the same series:
- High with high: nhỏ nhắn (hỏi + sắc), lửng lơ (hỏi + ngang).
- Low with low: vững vàng (ngã + huyền), mạnh mẽ, lặng lẽ, đẹp đẽ (nặng + ngã).
So if you are unsure whether lẽ in lặng lẽ takes hỏi or ngã, the partner lặng (low series) points to ngã. It is not a perfect rule, but it is the trick many Vietnamese use to get the hỏi/ngã spelling right without being able to hear the difference.
What I’d tell a learner
- Pick one dialect and copy its whole package. Northern tones include voice quality; Southern tones are mostly pitch with hỏi and ngã merged. Mixing them sounds odd to both sides.
- If you learn the northern standard, practise the creak. Imitate ngã and nặng from native recordings and listen for the catch in the voice, not just the pitch line.
- Learn tones as part of the word. Ga (station) and gà (chicken) are as different as “bat” and “bit”. Memorise every new word with its tone from day one.
- Use the high/low pairing for spelling. In reduplicated words, the tone of one half usually tells you which series the other belongs to.
- Don’t panic about Southern speech. If sửa and sữa sound identical to you in Sài Gòn, your ears are fine; that is the dialect.
Sources
- Kirby, J. (2011). Vietnamese (Hanoi Vietnamese), Journal of the International Phonetic Association 41(3), 381–392.
- University of Alberta Phonetics Lab, Tone and voice quality in Hanoi Vietnamese.
- Brunelle, M. (2009). Tone perception in Northern and Southern Vietnamese, Journal of Phonetics 37(1), 79–96 (author’s page).
- Jewell (2022), Yale BA thesis on Southern perception of the hỏi–ngã contrast.
- Tuổi Trẻ News, Learning Vietnamese: the alphabet and phonetics.
- Haudricourt, A.-G. (1954). The origin of tones in Vietnamese (English translation of De l’origine des tons en vietnamien).
- Thurgood, G. (2002). Vietnamese and tonogenesis: revising the model and the analysis, Diachronica 19(2), 333–363.