Chinese Tone Pairs | HSK Study Notes

You can name all four tones, you can read the marks, and you still cannot tell whether the speaker just said 白 (bái, “white”) or 百 (bǎi, “hundred”). This is the normal middle stage of learning Mandarin, and it is not a memory problem. Tone marks are knowledge; hearing tones is a trained perceptual skill, and it is trained by comparison, not by recitation.

The tool for that comparison is the minimal pair: two words with the same syllable and different tones, so the only thing left to hear is the pitch. This guide builds a drill you can run in two minutes a day, in the order that actually pays off.

The short version: learn the shape of each tone rather than its number, drill second versus third tone before anything else, always force yourself to choose between exactly two options, and score the drill so you know when to move on.

First, what you are listening for

A tone is a pitch movement across one syllable. Chinese phonetics describes these movements on a five-level scale, where 1 is the bottom of your comfortable range and 5 is the top. Reading the numbers as a path is more useful than memorizing a label.

Tone	Pitch path	What it does	Example
First	5 → 5	stays high and flat, no drift down	妈妈 (māma, “mum”)
Second	3 → 5	starts mid and climbs to the top	麻烦 (máfan, “trouble; troublesome”)
Third	2 → 1 → 4	dips low, holds low, then rises at the very end	马 (mǎ, “horse”)
Fourth	5 → 1	drops from the top, fast and decisive	骂 (mà, “to scold”)
Neutral	short, light	no shape of its own, borrows from the word before	the second syllable of 妈妈 (māma)

Two things in that table matter more than the rest.

The third tone’s rise arrives late, and often not at all. In a dictionary or a slow drill, 马 (mǎ) dips and then rises. In connected speech, a third tone before any non-third tone is usually just low — you drop and stay there. That is the half third tone, and it is the version you will hear far more often.

The fourth tone is defined by speed, not by volume. Learners tend to shout it. It is a fast fall from high, and it is shorter than the other three. If you make it loud instead of quick, it starts to sound like a first tone that ran out of breath.

What English gives you here, and what it breaks

Start with the good news, because it is real: English prosody already distinguishes words. A REcord is not the same thing as to reCORD; a PREsent is not to preSENT; an INsult is not to inSULT. Same consonants, same vowels on the page, different words. You do not need convincing that something beyond the consonants and vowels can decide which word you just heard.

What English does not give you is the kind of signal Mandarin uses. English stress is carried by loudness, length, and vowel quality; a Mandarin tone is a pitch direction. What separates the tones is not how hard or how long the syllable is, but whether the voice rises, falls, or holds. So the most common substitution is exactly the one you would expect: “move the pitch” quietly becomes “say it louder”, and a firmly stressed but flat syllable is not a tone at all — it is a first tone with emphasis.

Three consequences worth watching for:

You hear a tone as an attitude rather than as part of the word. English reserves pitch for sentence meaning: statements fall, questions rise, and a sharp fall signals emphasis or irritation. So a fourth tone sounds curt, a second tone sounds like a query, and because your ear files all of that under “tone of voice” — a revealing phrase in itself — it discards the information the moment you concentrate on meaning. This is why you can be accurate in a vocabulary drill and lose the tones completely in a sentence. The fix is a decision rather than an exercise: in Chinese the tone is part of how a word is spelled. 骂 (mà, “to scold”) is not an annoyed 妈 (mā); it is a different word.
Sentence-final falling intonation overwrites the last syllable. English declaratives drop in pitch at the end, so a first tone or a second tone in final position gets dragged down and flattened. Record the same word alone and then at the end of a sentence, and compare — this one is easy to catch and easy to miss.
Unstressed syllables get reduced to a schwa. English swallows the vowels it does not stress, and a swallowed vowel takes its tone with it, because a tone needs a vowel to run along. Every Mandarin syllable keeps its full vowel and its own tone. The only exception is the genuine neutral tone, and that is a property of the word, not something you get to apply wherever a syllable feels unimportant.

Why second and third tone is the pair to fix first

Of the six possible tone pairings, one causes most of the trouble for most learners, and it is second versus third. Both of them end with the voice going up, so if you listen to the end of the syllable — which is the instinctive thing to do — the two sound nearly identical.

The discriminator is at the beginning:

Second tone starts around the middle of your range and moves up immediately. There is no low part.
Third tone starts low and stays low. Any rise comes late, and in a sentence it usually never comes.

So retrain the question you ask yourself. Not “did it go up?” but “did it start low?” If the syllable begins in the basement, it is a third tone, whatever happens afterwards.

Same syllable	Second tone	Third tone
bai	白 (bái), “white”	百 (bǎi), “hundred”
wan	完 (wán), “to finish”	碗 (wǎn), “bowl”
yin	银行 (yínháng), “bank”	饮料 (yǐnliào), “drink; beverage”
xue	学习 (xuéxí), “to study”	雪 (xuě), “snow”
shi	时候 (shíhou), “time; moment”	使 (shǐ), “to use; to cause”

Do not try to learn all five rows at once. Take the first row, get it reliable, and add the next one tomorrow.

The other pairings, in order of usefulness

Once second versus third is stable, the remaining contrasts fall quickly, because each one differs in two features at once rather than one.

Contrast	The cue that separates them	Practice pair
Third vs fourth	third stays low, fourth falls from high	买 (mǎi), “to buy” / 卖 (mài), “to sell”
First vs second	first is flat, second climbs	汤 (tāng), “soup” / 糖 (táng), “sugar; candy”
Second vs fourth	opposite directions — up versus down	十 (shí), “ten” / 是 (shì), “to be”
First vs fourth	first holds high, fourth leaves high	书 (shū), “book” / 树 (shù), “tree”
Third vs first	third is low, first is high and flat	洗 (xǐ), “to wash” / 西 (xī), “west”

A few more verified pairs to rotate through once those feel easy: 好 (hǎo, “good”) and 号 (hào, “number; date”); 九 (jiǔ, “nine”) and 就 (jiù, “then; precisely”); 狗 (gǒu, “dog”) and 够 (gòu, “enough”); 茶 (chá, “tea”) and 差 (chà, “poor; lacking”); 猫 (māo, “cat”) and 毛 (máo, “hair; fur”); 醒 (xǐng, “to wake up”) and 姓 (xìng, “surname”).

The four-way set, only after the two-way sets work

The classic textbook quartet — 妈 (mā, “mother”), 麻 (má, “hemp”), 马 (mǎ, “horse”), 骂 (mà, “to scold”), shown here as an illustration rather than as a vocabulary list to memorize — looks like the obvious starting point, and it is the worst place to start. Choosing between four options at once is a harder task than hearing a difference, so a failure tells you nothing about which contrast you are actually missing.

Use them as a check, not as training. When you can pass a four-way set at speed, without the words in a fixed order, that contrast is genuinely in your ear. Complete four-way sets made of common words are rarer than textbooks suggest, but this one is entirely official HSK vocabulary:

tang: 汤 (tāng, “soup”) — 糖 (táng, “sugar”) — 躺 (tǎng, “to lie down”) — 烫 (tàng, “scalding hot”)

A three-way set works just as well as a check, and there are more of them to choose from: 包 (bāo, “to wrap; bag”) — 饱 (bǎo, “full, of having eaten enough”) — 抱 (bào, “to hold; to hug”).

The two-minute drill

The drill has to be forced-choice: you commit to an answer before you see it. Recognition without commitment feels productive and trains almost nothing.

Pick one pair. One, not five. Say 白 (bái) and 百 (bǎi).
Listen and commit. Play one of the two at random and say out loud which it was, before checking. Ten trials.
Say each one three times, and record the third. Recording is the only way to hear the gap between what you intended and what you produced.
Check exactly one feature on playback. For this pair: did the syllable start low, or start mid? Nothing else.
Put it in a phrase. 白色 (báisè, “white; the color white”) and 一百 (yìbǎi, “one hundred”). A tone that only survives in isolation is not usable yet.

Where to get the audio matters less than that it is a native voice and that you can replay one syllable on demand. Recordings in a course, a dictionary app with audio, or the audio in HSK listening material all work. What does not work is reading the pinyin aloud and grading yourself — that only tests whether you can read your own guess.

How to score it, and when to move on

Give yourself a real threshold instead of a feeling. Eight out of ten forced-choice trials, two days in a row, on a pair you did not practice immediately beforehand. Below that, keep the pair. At or above it, retire the pair and take the next one.

Two failure patterns worth naming, because they need different fixes:

You hear it but cannot say it. Your perception is ahead of your production, which is the normal order. Fix it with recording and exaggeration: make the movement two or three times bigger than feels natural, then shrink it back once it is reliably audible.
You can say it but cannot hear it. Usually you are listening to the end of the syllable. Go back to the “did it start low?” question, and slow the audio down rather than repeating it at full speed.

Do not stop at single syllables

Single-syllable drills are scaffolding. Real speech gives you tones in sequence, where each one is colored by its neighbors, so a learner who is perfect on isolated syllables can still lose the thread in a sentence.

Two things change as soon as you combine syllables. The first is tone sandhi: two third tones in a row turn the first one into something close to a second tone, so 你好 (nǐ hǎo, “hello”) is said roughly as ní hǎo. The second is the neutral tone, which absorbs the shape of whatever came before it — 妈妈 (māma) is not two first tones, it is a first tone with a light tail.

Words that contain a tone contrast inside themselves are the best bridge, because you cannot fake your way through them: 牛奶 (niúnǎi, “milk”) is a second tone followed by a third, and 米饭 (mǐfàn, “cooked rice”) is a third followed by a fourth. Say those and you are already doing the real thing.

When you are ready, move on in this order:

Chinese tones explained for the four shapes and how they are written.
Chinese tone sandhi for the third-tone change and the shifting tones of 一 (yī) and 不 (bù).
Chinese tone phrase drills for two-syllable and sentence-level practice.
Neutral tone and erhua for the unmarked fifth tone.
Chinese listening training to put the trained ear to work on longer audio.

What this buys you in HSK listening

HSK listening does not test tones directly, which is exactly why tone discrimination decides your score. The distractors in the lower levels are frequently built from words that share a syllable and differ in tone, so a shaky contrast turns a question you understood into a coin flip. Higher up, the cost is different and larger: every syllable you have to stop and resolve is a syllable you are not spending on the sentence, and the audio does not wait.

The practical consequence is that tone drilling belongs in your listening practice, not in a separate pronunciation slot. When you miss a word in a listening exercise, check whether you misheard the tone or genuinely did not know the word. Those are different problems, and only one of them is fixed by learning more vocabulary.

Summary

Learn each tone as a pitch path, and remember that the third tone’s rise usually does not arrive in real speech.
Fix second versus third tone first, and use the start of the syllable as your cue, not the end.
Drill in forced-choice pairs of two, and keep four-way sets as a test rather than as practice.
Record yourself, check one feature per playback, and exaggerate before you refine.
Move on at eight out of ten across two days, then take the contrast into two-syllable words and sentences.

LISTENING MINI DRILL

Listening mini-drill: choose the tone

Press play, cover the spelling, and choose the contour you hear. This uses a Chinese browser voice when available; use a clear learning-source recording for a final comparison.