You already know the first tone is high and flat, the second rises, the third dips, the fourth falls. Knowing that has not fixed your tones. What fixes them is a program — Chinese tone practice exercises you run in a fixed order, every day, with a stated goal and a way to tell whether you passed. Most learners never get one. They re-read the theory, feel vaguely worse, and go back to vocabulary.
What follows is a workout plan: seven exercises, each with a goal, a method, a duration, and a pass criterion, plus an eight-week schedule. If tones are still new to you, start with the beginner's guide to learning Chinese tones and the four tones with audio examples, then come back.
Why 15 minutes daily beats 90 minutes on Sunday
Tone production is a motor skill with a perception loop attached — closer to a tennis serve than to a vocabulary list. Two things follow.
Consolidation happens between sessions, not inside them. Six 15-minute sessions across a week give you six cycles of practice, sleep, and re-testing. One 90-minute session gives you one, and by minute 40 you are rehearsing a tired voice and reinforcing the mistakes fatigue creates.
Error correction needs fresh attention. A drill pays off in the moment you produce a tone, hear it was wrong, and try again — and that loop degrades fast once you drift. So: 10–15 minutes, daily, same slot. Miss a day and just resume.
The seven Chinese tone practice exercises, in order
Everyone asking how to improve Chinese tones eventually lands on the same answer: not more theory, more reps with feedback. Run these in order — each assumes the previous one is roughly working.
1. Tone pair drills — the exercise almost everyone skips
Goal: produce all 20 two-syllable tone combinations correctly. Isolated tones are easy; combinations are where learners fail, because the difficulty is the transition — resetting your pitch between syllable one and syllable two.
How to do it: work the table below one row at a time. Say the word, record it, check both syllables scored correct, and do not advance until the current row passes. Order matters: start with the maximally distinct pairs (1+4, 4+1), then the same-tone pairs, and save the second/third tone family (2+2, 2+3, 3+2, 3+3) for last — that cluster is where tone practice in Mandarin quietly falls apart.
| Combination | Word | Pinyin | Meaning |
|---|---|---|---|
| 1 + 1 | 飞机 | fēijī | airplane |
| 1 + 2 | 中国 | zhōngguó | China |
| 1 + 3 | 铅笔 | qiānbǐ | pencil |
| 1 + 4 | 工作 | gōngzuò | work |
| 1 + neutral | 妈妈 | māma | mom |
| 2 + 1 | 明天 | míngtiān | tomorrow |
| 2 + 2 | 学习 | xuéxí | to study |
| 2 + 3 | 苹果 | píngguǒ | apple |
| 2 + 4 | 学校 | xuéxiào | school |
| 2 + neutral | 朋友 | péngyou | friend |
| 3 + 1 | 老师 | lǎoshī | teacher |
| 3 + 2 | 美国 | měiguó | United States |
| 3 + 3 | 你好 | nǐhǎo | hello |
| 3 + 4 | 好看 | hǎokàn | good-looking |
| 3 + neutral | 姐姐 | jiějie | older sister |
| 4 + 1 | 汽车 | qìchē | car |
| 4 + 2 | 面条 | miàntiáo | noodles |
| 4 + 3 | 电脑 | diànnǎo | computer |
| 4 + 4 | 再见 | zàijiàn | goodbye |
| 4 + neutral | 谢谢 | xièxie | thank you |
One caveat on the 3 + 3 row: 你好 is written nǐhǎo but spoken with the first syllable raised, closer to ní hǎo. That is a rule, not sloppiness — learn the third tone sandhi rules before drilling that row.
How long: 5–7 minutes, roughly four combinations per session.
How you know you passed: 8 of 10 consecutive attempts correct on both syllables, achieved on two different days. Passing once is luck.
2. Minimal pair production
Goal: control tone independently of the syllable. Same sounds, different tone, on demand.
How to do it: produce a minimal set in order, then in random order. The random pass is the real test — in sequence you are riding momentum, not choosing a tone.
- mā má mǎ mà — 妈 麻 马 骂 (mom, hemp, horse, to scold)
- tāng táng tǎng tàng — 汤 糖 躺 烫 (soup, sugar, to lie down, scalding)
- wén 闻 vs wèn 问 (to smell / to ask)
- mǎi 买 vs mài 卖 (to buy / to sell)
- shū 书 vs shǔ 鼠 (book / mouse)
The buy/sell pair deserves extra reps: it is the clearest demonstration that a tone error is not an accent, it is a different word.
Build your own sets from the interactive pinyin chart — pick a syllable, hear all four tones back to back, then produce them yourself.
How long: 3 minutes.
How you know you passed: the randomized order scores as clean as the sequential order. If random is worse, you were reciting, not producing.
3. Exaggeration, then normalization
Goal: build the pitch contour big, then shrink it to conversational size without losing the shape.
How to do it: three reps of each word wildly exaggerated — a pitch range far wider than any real speaker uses. Three at medium. Three at natural conversational range. Same word, same session.
Exaggeration first is correct because the pitch movement you think you are producing is always smaller than what you intend; overshooting builds a motor target you can actually feel. But it is a scaffold, not a destination. Learners who stay there develop the sing-song delivery natives call "textbook Chinese" — every syllable equally stressed, contours too wide, no rhythm. The drill is finished only when the natural version still scores correct.
How long: 2–3 minutes, folded into exercises 1 and 2.
How you know you passed: the natural-range rep scores the same tone as the exaggerated rep. If it does not, your contour is not stable yet — it is just loud.
4. Shadowing
Goal: transfer tones into connected speech at native timing.
How to do it: pick a short clip (20–40 seconds) with a clear speaker. First pass: audio only, no text, echoing about half a second behind. Second pass: same clip with the transcript in front of you. Third pass: audio only again.
Two things make shadowing fail:
- Audio above your level. If you cannot identify most of the words, you are copying noise. Use material where you know 90% of the vocabulary.
- Reading instead of listening. Once text is in front of you, most learners stop hearing the model and read aloud with their own tones. That is why the audio-only pass comes first and last.
How long: 3–4 minutes.
How you know you passed: three run-throughs with no dropouts, and your recorded final pass scores the same tones as the source text.
5. Sentence-level tone maintenance
Goal: transfer. Tones survive drills and die in sentences; this is the exercise that catches it.
How to do it: read a sentence aloud, record it, and check the tones syllable by syllable — not "did it sound okay". Start with these:
- 我想买一本中文书。 Wǒ xiǎng mǎi yì běn Zhōngwén shū. (I want to buy a Chinese book.)
- 他明天去北京开会。 Tā míngtiān qù Běijīng kāihuì. (He is going to Beijing tomorrow for a meeting.)
- 你们的老师是美国人吗？ Nǐmen de lǎoshī shì Měiguórén ma? (Is your teacher American?)
The first sentence is deliberately cruel: wǒ xiǎng mǎi is three third tones in a row. The typical failure pattern is that tones flatten mid-phrase, third tones stop dipping, and fourth tones lose their fall because you are rushing the end.
Once these are clean, use your own material — practice any Chinese text with AI feedback and paste in sentences from whatever you are actually studying. Drilling sentences you will never say is how tone work becomes a hobby instead of a skill.
How long: 3–5 minutes, five sentences.
How you know you passed: the syllables that pass in isolation also pass inside the sentence. The gap between those two numbers is the most useful measurement in this program.
6. Dictation — the reverse drill
Goal: train perception, which sets the ceiling on production. You cannot reliably produce a contrast you cannot hear.
How to do it: have an app or a partner play single syllables. Write the tone number, 1 to 4, or 0 for neutral. Twenty per round, weighted toward second and third tone — the confusion that survives longest. Then invert it: play a two-syllable word and write both numbers. Harder and far more useful, because it is exercise 1's transition problem heard instead of spoken.
For the full perception method, see how to train your ears to distinguish Mandarin tones; this drill is its daily maintenance version.
How long: 2–3 minutes.
How you know you passed: 18 of 20 correct on syllables you did not choose yourself, including at least six 2/3 items.
7. Spaced review of your own failures
Goal: spend your reps on what you actually get wrong.
How to do it: keep an error log — one line per miss: the syllable, the tone you produced, the tone you meant, the date. Re-test each entry after 1, 3, 7, 14 and 30 days. Three clean passes and it graduates off the list.
Most people re-drill what already works because it feels good. Your log will be shorter and stranger than you expect — usually a handful of syllables, clustered in one or two tone-pair combinations, accounting for most of your misses.
One distinction: this is about producing tones you keep missing. Remembering which tone a word carries is a separate problem — see how to memorize Chinese tones. For why the intervals are spaced rather than massed, spaced repetition for Chinese pronunciation explains the mechanism.
How long: 2 minutes, and more valuable every week.
How you know you passed: your log is shrinking faster than it grows.
An eight-week schedule
Print this. Each week keeps the previous week's drills as a short warm-up.
| Week | Focus | Daily minutes |
|---|---|---|
| 1 | Baseline recording + minimal pairs (exercise 2) | 10 |
| 2 | Tone pairs starting with tone 1 and tone 4 (1+x, 4+x) | 12 |
| 3 | Tone pairs starting with tone 2 and tone 3 (2+x, 3+x) | 15 |
| 4 | Neutral-tone pairs + exaggerate-then-normalize | 15 |
| 5 | Shadowing, audio only, plus tone-pair failures | 15 |
| 6 | Sentence-level maintenance, five sentences a day | 15 |
| 7 | Dictation + spaced review of the error log | 15 |
| 8 | Re-record the baseline, rebuild the log, practice your own texts | 15 |
Two months of that is roughly seven hours of focused work. It will do more for how you sound than a year of passive listening.
How to measure progress honestly
Your own ear is the worst instrument for judging your own tones, because you hear what you intended to say. Replace feel with four numbers:
- A frozen re-test set. Twenty words you never drill, recorded in week 1 and again in week 8. Improvement there is real transfer, not memorized reps. A quick Chinese pronunciation test works as a standing baseline.
- Per-syllable tone accuracy. A score that tells you which syllable failed, not just that the word was wrong. Vague pass/fail feedback cannot drive a drill program.
- The isolation-to-sentence gap. Score the same 10 words alone and inside sentences. A shrinking gap means your tones are becoming automatic.
- Error log velocity. Entries added per week versus entries graduated per week.
None of these needs a teacher, and all are more honest than "I think it's getting better." For why automated feedback closes this loop faster than self-assessment, how to fix your Mandarin tones with technology makes the case.
Start today, not on Monday
The first week of Chinese tone drills feels humiliating. You will find combinations you assumed were fine — usually in the 2 and 3 family — that have been wrong for months. That discovery is the point. Everything after it is reps.
TonePerfect scores your Mandarin syllable by syllable, splitting each one into initial, final, and tone, so you can run every exercise above and know exactly what failed. It works in your browser, no install — try it free, or open the pinyin chart and pick your first minimal pair. Fifteen minutes today beats a perfect plan you start next month.