How to learn Chinese tones as an adult: what research shows

Tones are hard for your memory, not your ear

English uses pitch across whole sentences, to ask a question or stress a word. So an adult English ear has never had to file a syllable's pitch as part of the word itself.

That makes the hard part listening and word memory, not your voice. The best evidence I found comes from people who got very far.

In one study, 16 English speakers with about 11 years of Mandarin identified tones on single syllables almost as well as native listeners. Then they judged whether syllables were real words. When only the tone was wrong, they usually missed it:

Source: Pelzl et al. 2019, Tables 1, 2 and 4. Naming tone 2: identifying the tone of a single tone 2 syllable. Spotting fakes: rejecting a made-up word that differs from a real one only in tone. On single syllables, the learners matched native listeners on every tone except tone 2.

They could hear tones. What they hadn't built was the habit of storing the tone as part of each word.

Nor are beginners' ears the obstacle. Adults with no Mandarin at all tell tones apart better than chance before any study, and they improve after a month of classes.

Tones aren't decoration you can skip, though. Identifying a syllable's tone carries at least as much information as identifying its vowel. And here's what happened when researchers flattened the pitch of Mandarin sentences:

Source: Patel, Xu & Wang 2010. In quiet, flattened sentences were understood about as well as natural ones.

Context rescues you at a quiet table, not in a busy restaurant.

You're probably not tone deaf. Fewer than 2 in 100 people have true tone deafness, and even they can learn word tones. See the tone-deaf section.

Four tones and a light one

Mandarin has four tones, plus a light, unstressed one called the neutral tone. The shapes in our lab describe a syllable said on its own.

In ordinary speech, the third tone usually loses its rise, and every tone bends toward its neighbours. Linguists write tone shapes with Yuen Ren Chao's pitch numbers, from 1 at the bottom of your voice to 5 at the top, as PolyU's introduction to phonetics explains.

For a printable chart and audio for every tone pattern, use the tone-pair trainer. To hear any syllable in all its tones, open the pinyin chart.

Train your ear first, but speak from week one

My advice: start with your ears, and say every word out loud from the first day. Listening training improves your pronunciation only partly, because each skill improves mostly through its own practice.

The listening half is well supported. In a 1999 study, eight American learners trained on tone identification for eight sessions over two weeks. Here's how their accuracy rose:

Source: Wang, Spence, Jongman & Sereno 1999.

The training reached the learners' voices too. After the same kind of listening-only practice, native listeners identified the trainees' own spoken tones 18% more accurately.

The "partly" matters. Across 25 years of studies, perception training had a large effect on listening, and one about half as big on speaking.

In a study of 38 English speakers learning tone words, people who practised only listening or only speaking did far worse on the skill they hadn't practised. So listen first, but say every word out loud too.

Hearing many different voices sounds like an obvious upgrade, and it can help you recognise new speakers later. But the evidence is mixed:

• A 2021 meta-analysis of speech-sound training found that the immediate advantage of many voices shrank to nothing, once outliers were removed and publication bias corrected. The advantages for handling new voices and for long-term retention stayed large.
• In another study, many voices helped learners who already perceived pitch well, and slowed down those who didn't.

If tones feel impossible, I'd start with one clear voice.

Try our ear test below. The default round pits tone 2 against tone 3, the pair that causes the most trouble. Guessing would get you 5 out of 10, so anything clearly above that is your ear working.

When single syllables feel easy, take the 1-minute ear quiz on all 20 two-syllable tone patterns, or play Panda Tone Toss.

Practise tones in two-syllable words

Tones change shape next to each other, and learners find them harder in two-syllable words than in single syllables. That's why I'd practise in pairs.

A tone is pulled toward the one before it. The end of one syllable strongly shapes the start of the next, while the following tone has a small opposite effect.

New learners in a training study learned tones less well in two-syllable words than in one-syllable words. Drilling shī shí shǐ shì one by one is a start. Saying tomorrow, 明天(míngtiān), and movie, 电影(diànyǐng), trains tones the way you'll actually use them.

Our grid below has one real word for each tone pattern. Blue dots mark pairs where tone 3 stays low; the coral one changes tone outright.

To drill every pattern with 20 words each, use the tone-pair trainer.

On paper, the free Tone-Pair Listening and Speaking Lab covers all 16 tone combinations plus light endings, in four 15-minute sessions.

Is the third tone low or dipping?

Usually low. It dips and rises only on its own or at the end of a phrase. Before tones 1, 2 and 4 it stays low, and before another third tone it rises like tone 2.

Phoneticians call the low version the "half third". Its pitch is about 21 instead of 214, as PolyU and Chen and colleagues describe it. We measured it in the lab's recordings:

Source: SayMei's computer voice, measured with the Praat pitch tracker. Both words play in the lab above.

Native speakers treat the low version as the normal case. They apply the low change to made-up words more reliably than the rising one. Learners lag behind: second-language speakers in the US used the low third tone before other syllables less often than heritage speakers did.

Part of the problem may be how the tone is taught. A thesis survey found that 14 of 15 beginner textbooks described tone 3 as falling and rising.

Researchers have argued that teaching the dip first may feed the tone 2 and tone 3 confusion, because tone 2 can dip slightly too.

What I'd do: aim for the bottom of your normal speaking range and stay there. Add the rise only at the end of a phrase. To see which third tones rise in a phrase you're learning, paste it into the tone sandhi checker.

Tone 2 and tone 3 are the pair to watch

Both can start low and rise. What separates them is when the pitch turns upward and how far it drops first, and even advanced learners find tone 2 the hardest to identify.

Listeners tell the two apart mainly by that turning point and that first drop. Tone 2 turns up early, after a small dip. Tone 3 sinks further and turns late, if at all. The advanced learners in the study above were near-native on every isolated tone except tone 2.

Even computer voices blur this line. We screened 16 one-syllable tone 2 recordings from SayMei's own audio library for the lab. Seven dipped too deep or turned up too late to count as clear examples, so the ear test leaves them out, as Methods explains.

Which tone is "hardest" depends on the task, so don't trust a single ranking:

• Listening: in a 2024 training study, tone 1 was easiest to identify and tone 3 hardest. Exaggerating the pitch range helped tones 2 and 3 but hurt tones 1 and 4.
• Speaking: American learners after four months of study made errors on 55.6% of tone 4 syllables but only 8.9% of tone 2 syllables, in one older study. Another, from the same year, found errors spread evenly.

The consistent trouble spot is the tone 2 and tone 3 pair. Two habits I'd build:

• Listen for the turn. In the lab's 2 vs 3 round, ask where the pitch bottoms out: near the start for tone 2, or past the middle for tone 3.
• Use more of your range when you speak. Native Chinese speakers use a pitch range about 1.5 times wider than English speakers do in English.

What is the neutral tone, and how do you say it?

A neutral-tone syllable is short and light, with no shape of its own. Its pitch follows the syllable before it: a little higher after a third tone, lowest after a fourth.

The neutral tone has a weak middle target that the previous tone overpowers. It's also about half as long as a full syllable, as Lin's review reports from earlier measurements.

Length is where learners go wrong. In one study, learners' neutral tones differed from native speakers' mainly because they were too long, not because of their pitch, and they got the lowest goodness ratings.

Sources agree that the light syllable sits low after tone 4, as in thank you, 谢谢(xièxie), and relatively high after tone 3, as in we, 我们(wǒmen). They disagree about its exact height after tone 1, so our lab draws that one as a range.

The light tone can change meaning. Thing is 东西(dōngxi), but 东西(dōngxī) means east and west, as CC-CEDICT lists them.

To practise, make the second syllable quick. Say mom, 妈妈(māma), then shorten the second "ma" until it nearly disappears.

Three tone changes to learn on day one

I'd learn three changes from the start:

• Two third tones in a row: hello, 你好(nǐ hǎo), is said ní hǎo.
• One, 一(yī), changes before a fourth tone.
• Not, 不(bù), changes before a fourth tone.

Dictionaries keep writing the original tones, so learn the spoken forms by ear. Here's how they sound:

Source: these match the rules in SayMei's tone sandhi checker.

With three or more third tones in a row, word grouping decides which ones rise. I'm fine, 我很好(wǒ hěn hǎo), can be said wǒ hén hǎo or wó hén hǎo. I'd leave that to the checker, not to memory.

Knowing the rule is the easy part. American learners with about three years of study applied the third-tone rules even to made-up words, but with less accurate pitch than native speakers.

For printable drills with dictionary and spoken tones side by side, get the free workbook Tone Changes in Everyday Mandarin.

Pitch graphs help; a teacher still hears more

Pitch graphs and teachers both help, in different ways. Seeing your own pitch curve improved tone accuracy in several studies. Automatic checkers judge single syllables imperfectly, and a teacher still catches errors in context.

We found four visual-feedback studies. Each is small, but they point the same way:

Source: the four studies, linked.

How you're corrected matters too. In a 14-week one-to-one course, beginners whose teacher simply repeated the word back correctly, a "recast", improved their spoken tones more than those given explicit correction. The study, by Bryfonski and Ma, had 41 learners.

What an automatic checker can hear is narrower than it looks. SayMei's free Tone Checker compares the pitch shape of one word with the model. In our own test, it named the intended tone 84% of the time:

Source: SayMei's internal evaluation file, 6 October 2026: 4,344 labelled clips from seven computer voices, from ElevenLabs and Gemini text-to-speech, plus pitch-shifted, flattened, slowed, sped-up and noisy copies.

A real third tone can dip or stay low, and that confuses the checker. One caveat matters more: every test voice was computer-generated, including the deeper, flatter and noisier altered copies. We haven't measured it on human learners yet, and we'll add those results when we have them. It also doesn't judge sentences, consonants or vowels.

My rule: apps for repetition, a human for diagnosis. If you have one session with a teacher, ask them to listen to the same ten words you practise every day.

Shadowing and gestures can help

One small study says shadowing works for tones. Shadowing means speaking along with a recording, a split second behind it.

In a study of 14 beginners, four weeks of shadowing improved tone accuracy in free speech, whether the recordings were authentic videos or textbook audio. That's a small sample, so treat it as promising rather than proven.

Hand movements help some people. In a study of 106 participants, watching and making pitch gestures, like a hand that rises for tone 2, improved tone identification and word learning in a short session.

To build up to sentences, our Tone-Pair Lab PDF has "ladders" that grow one tone pair into a phrase and then a full sentence, with free audio. For a timed speaking drill you can do alone, see How to speak Chinese.

Your ear improves in weeks; your words take years

Your ear can improve within days or weeks. Using tones automatically in real words takes years. Here's how fast each kind of progress came in the studies we read:

Source: the four studies, linked.

The slow part is recognising words. The gap between advanced learners and native listeners persisted even in near-ideal listening conditions.

What no study shows. "Most learners master tones in two to three months" appears in blog posts and chatbot answers, with no source. I couldn't find a study that supports it, and the advanced-learner results point the other way. Expect clear progress in weeks, and a long tail of practice in real conversations.

You're almost certainly not tone deaf

True congenital amusia, the clinical form of tone deafness, is rare, and even people who have it learned new word tones in a training study.

The old figure of 4% came from a single 1980 estimate.

A test of 20,000 people put it at about 1.5%. Amusia does make tones harder, though:

• Among 22 Mandarin speakers with amusia, nearly half had trouble telling tones apart. The six with marked difficulty still produced tones normally.
• In a training study, 21 people with amusia were less accurate than 23 typical listeners, but they still learned new word tones.

Musical ability helps, but you don't need it. Pitch perception and musical experience predicted who learned pitch-based words best, with large individual differences.

Musical ability also went with tone accuracy in the listening-or-speaking study. And pitch acuity predicted tone identification in Portuguese-speaking learners.

If you struggle, it usually means more listening practice, not a missing ability.

A six-week plan for adult beginners

Ten to fifteen minutes a day is enough to start. That's our suggestion, not a study finding. Each week adds one skill, and every step uses one of our free tools:

Keep a fixed list of ten words, and re-test yourself once a week, for listening and speaking separately. A fixed list makes the scores comparable.

Six tone myths the evidence doesn't support

Each of these shows up in guides and chatbot answers. Here's what the studies actually found:

The pattern is the same in each row: the confident version is simpler than the evidence.

Where SayMei fits, and where it doesn't

Our free tools cover listening and single-word checking, and Mei Lin gives you spoken practice. None of it replaces a teacher's ear for persistent errors.

Here's our free tool for each step, with no sign-up:

Where it fits

• The free first lesson puts your tones under conversational load, which is where the research says they break. It runs up to 10 minutes, with no account or card.
• After that, SayMei Premium is $5.99 a month for 200 minutes with Mei Lin. It starts with a 7-day trial with 100 minutes that needs a card; see our pricing page.
• The HSK 1 songs help with words and listening. They aren't tone models; see the FAQ.

Where it falls short

• No human ear. A trained teacher hears register errors, too high or too low, and mistakes in context that SayMei's tools miss.
• The Tone Checker is narrow. It checks one word at a time, tones 1 to 4 only. Its accuracy comes from computer voices, and it's weakest on tone 3. Its earlier version, replaced in October 2026, recognised our own reference recordings on only 50 of 96 cards.
• Mei Lin's ability to catch tone errors hasn't been measured. In the first lesson she's instructed to correct by repeating your sentence back the right way, one fix at a time. That's the style that worked in the recast study above, but that study used human teachers. Treat her as practice, not diagnosis.
• The audio is synthetic and uses few voices. The lab, chart and trainer use SayMei's text-to-speech. Real speakers vary more.

Skip SayMei for tones if you need a teacher to diagnose one persistent error, or detailed feedback on whole sentences. Book a human tutor for that.

What to do next

• Test your ear today. Run the 2 vs 3 round in the lab, then the tone-pair quiz.
• Say every word out loud. Say it from the written prompt first, then check it against the audio.
• Keep the third tone low. Add the rise only at the end of a phrase.
• Learn the three day-one changes. Two third tones in a row, one, 一(yī), and not, 不(bù).
• Get a human ear on your list. Ask a teacher to check the same words you practise every day.