The Four Mandarin Tones: What Each One Sounds Like

Everybody ships the same five lines

Search for the four tones and you will get the same picture a hundred times: a grid, five pitch levels, four coloured lines. Flat, rising, dipping, falling. It is accurate and it is where almost every learner starts, and about a month later the same learner discovers that real Mandarin does not seem to contain any of those shapes.

The chart is not lying. It is drawing the citation form - the shape a tone takes when a syllable is said alone, slowly, by someone who has been asked to demonstrate a tone. That is a real thing, and it is roughly one per cent of the Mandarin you will ever hear. Everything else is faster, shorter, and reshaped by the syllables on either side.

What follows is the other description: not the shape each tone is drawn as, but the thing you are actually listening for when a word goes past at speed. Tingyin is a Mandarin tone-listening trainer - a clip plays, you pick which tone you heard, it tells you at once - so this is written from the identification side rather than the pronunciation side. Level 1 is free and needs no account, which is the cheapest way to find out whether any of this transfers to your ear.

First tone: level, and it refuses to drift

The first tone sits high in the speaker's range and stays there. Linguists write it 55 on a five-point scale: starts at five, ends at five. The description everyone gives is flat, and that word does the tone a disservice, because flat sounds like the default and it is anything but.

English does not hold a steady pitch on a syllable. Say the word tea on its own and listen: the pitch drifts, usually downward, because English syllables carry the tail end of a phrase contour. Holding a note dead level for the length of a syllable is a deliberate act, and that is the clue. The cue that identifies a first tone is not its height and not its length, both of which sit in the middle of the pack: it is the refusal to move. Across the 280 single-syllable clips Tingyin plays at levels one and two, first tones average 1.05 seconds against 1.01 for fourth tones, a gap of 40 milliseconds that nobody can hear.

妈 mā mother, 书 shū book, 天 tiān sky, 高 gāo tall. If you are guessing on a drill and you hear something that seems to have no pitch event in it at all - no lift, no drop, no scoop - that is usually a first tone rather than a syllable you missed.

Second tone: it rises from the middle, not from the bottom

This is the tone that a diagram most reliably mistrains, because the arrow points from bottom left to top right and learners duly listen for something that starts at the floor. It does not. The second tone is written 35: it starts around the middle of the speaker's range and rises to near the top. The rise is real but it is short and it happens in the upper half of the voice.

That detail matters more than almost anything else in this article, because it is the whole of the second-against-third confusion. If you are listening for "starts low" you will hear a third tone and call it a second, and hear a second tone and wait for a low start that never comes. The reliable contrast is not rising versus dipping. It is where the syllable begins: the second starts mid, the third starts low and stays there.

The second complication is English. A rise across a word is how English marks a question, so a second tone in the middle of a sentence gets routed to the part of your brain that handles sentence attitude, and the word identity is lost on the way. This is covered at length in why Mandarin tones are hard to hear, and it is the single most common way a fluent-sounding beginner turns out not to be following the conversation.

麻 má hemp, 人 rén person, 来 lái to come, 学 xué to study.

Third tone: usually just low

The third tone is drawn as a dip: down to the bottom, then back up. Written out it is 214 - start at two, fall to one, rise to four. That contour exists. You will hear it when a third-tone syllable is said alone, and at the end of a phrase where it has room to finish. Everywhere else it does something simpler.

Before another syllable, a third tone drops to a low flat pitch and stops. No rise. What identifies it is that it is the lowest thing in the phrase, often with a creak or a rasp in the voice as the pitch bottoms out - that creak is a better cue than any contour, because it is a physical consequence of speaking at the very bottom of your range and it survives fast speech.

And before another third tone it stops being a third at all: it is pronounced as a second. Tingyin's corpus labels the tone of every syllable as spoken rather than as written, and across all 637 clips there is not one tagged third-then-third. Words like 水果 shuǐ guǒ and 可以 kě yǐ are in the corpus; they are simply stored as second-then-third, because that is what comes out of a mouth. The full story is in why the textbook dip is the third tone's rarest sound and in Mandarin tone sandhi.

马 mǎ horse, 我 wǒ I, 好 hǎo good, 水 shuǐ water.

Fourth tone: short and sharp

The fourth tone falls from the top of the range to the bottom, written 51, and the thing nobody tells you is how fast. In running speech the fall is often too quick to perceive as a movement at all - what you register is a syllable that arrives loud and high and is gone.

It is the shortest of the four on average, but only just, and only against the third. In isolation it is a hair shorter than a first tone and about a quarter shorter than a third, so length is a cue for the third tone and close to useless for this one. What does separate a fourth is the attack rather than the duration. If a syllable lands hard and ends before the ones around it would have, that is a fourth. English speakers consistently report that fourth tones sound angry or emphatic, which is the same misrouting as the second tone: a sharp fall is how English marks finality and force, so the pitch gets read as an emotion rather than as part of the word.

骂 mà to scold, 是 shì to be, 大 dà big, 看 kàn to look. It also makes the fourth tone the easiest of the four to identify in a drill, which is worth knowing when your accuracy on it runs ahead of everything else and you wonder why the average has not moved.

The fifth one, which is not a tone

Most tone charts add a fifth row for the neutral tone, and it does not belong there. The neutral tone has no contour of its own - it is a short unstressed syllable that takes its pitch from whatever came immediately before it, which means it lands in four different places depending on the neighbour. It appears in 80 of Tingyin's 637 clips and never in the single-syllable levels, for the obvious reason that a syllable with no preceding syllable has nothing to borrow from.

The practical advice is to treat it as a rhythm rather than a pitch: it is the short, quiet, swallowed one. Where it lands after each of the four tones, and why it gets confused with the third, is the subject of the neutral tone in Mandarin.

A diagram cannot teach a sound

Everything above is still description. It is better description than the five-line chart, and it will not by itself let you identify a tone you hear. Perceptual categories are built from exposure with feedback, not from accurate accounts of what to expect - which is why someone can recite the contours perfectly and still fail a listening test the same afternoon.

Two things are worth taking from this into whatever you practise with:

• Identification is a different skill from production. You can learn to say a passable second tone in a week by copying a model. Telling a second from a third in half a second of unexpected speech takes months, and nothing about the first skill delivers the second.
• Drill the pairs, not the set. Second against third is where the time goes. First and fourth are comparatively easy and practising all four together spends most of your attention on contrasts you already have.

Tingyin's first level is exactly that exercise: 80 single-syllable clips built as minimal pairs, the same syllable in each of its tones, so the only variable is the one you are trying to hear. Level 2 adds 200 more single syllables, level 3 moves to 237 two-syllable words and level 4 to 120 longer ones where the sandhi and the neutral tones live. All 637 are human recordings - 329 from Yue Tan, 279 from Chen Wang and 29 from named Lingua Libre volunteers - and a word only makes it into a level if a recording of it exists carrying its source, speaker, licence and a content hash, so the vocabulary follows the audio rather than the other way round. Nothing measures the pitch of a shipped clip; the check confirms the bytes you are served are the bytes that were recorded.

Level 1 runs without an account and level 2 is free once you sign in. Levels 3 and 4 are paid, along with progress sync across devices and offline downloads: on the web, which is the only place to pay while the native apps are unpublished, that is $2.49 a week, $5.99 a month, or $15.99 once. If you would rather not pay anything, Anki with a deck of recorded pairs does the same job for the cost of building the deck yourself.

Frequently asked questions