Mandarin runs on about four hundred distinct syllables, and roughly thirteen hundred once the tones are counted. Every word in the language is built from that small set, which means an enormous number of words sound the same or nearly the same, and a listener is separating them by context rather than by sound.
So listening here is not a matter of catching the syllable more accurately. It is a matter of knowing which of the several words that syllable could be is the one this sentence needs, and that is prediction rather than perception.
Six things about what you will actually hear
- there are few syllables and many homophones. Around four hundred syllables ignoring tone, so a great many words sound identical, and this is precisely why two character compounds became the norm.
- context carries the meaning more than in any European language on this list, which means listening improves in large steps as vocabulary passes thresholds rather than smoothly.
- erhua in the north, an r sound attached to the end of syllables, heavy in Beijing and almost absent in the south, and it changes the shape of familiar words.
- 那个 as a filler, meaning that one, used the way English uses um. It is extremely frequent and is worth recognising early so it stops being parsed as content.
- 就是, meaning exactly or precisely, used as a strong agreement.
- 对对对, right right right, repeated, which is the ordinary way of agreeing and not a sign of impatience.
Tone perception lags production by about a year
Learners are usually able to produce the four tones long before they can hear them at speed, which is the reverse of the order in most languages and surprises people. The reason is that a produced tone is made deliberately, in your own range, with as much time as you want, and a heard tone arrives compressed, relative to a speaker whose range you do not know, and altered by the tones around it.
Two of those alterations are systematic and worth expecting rather than discovering. Third tones change before other third tones, so what you hear is not what is written. Many second syllables are neutral, carrying no tone at all, so a two syllable word arrives with one tone rather than two. A listener waiting for four clean contours will not find them, because they are not there.
What replaces the waiting is the same thing a native listener does. The topic, the sentence so far and the other syllable of the compound narrow the candidates to one, and the tone confirms the guess rather than making it. That is why a familiar subject at full speed is easier than an unfamiliar one slowed down.
Every listening failure gets called speed
Here is the stall and it is a misdiagnosis. Somebody fails to follow a conversation, concludes the speaker was too fast, and slows the recording down. Comprehension does not improve, because the missing information was never in the speed: the syllables were caught and the words could not be selected from them.
The test that separates the two takes a minute and it is worth running before another month of listening practice. Play a difficult clip at three quarters speed. If it becomes clear, the rate was the problem and exposure is the answer. If it stays opaque, read the transcript: if the transcript is easy, the words were known and the problem is that they were only ever known on the page, and the repair is meeting them in audio deliberately. If the transcript is also hard, it is vocabulary. Three outcomes, three different months of work, and the whole diagnosis costs less than an hour, which is the argument in the stalls every learner hits, and the way through each.
Two banks
From tone drills to a telephone call
Every plank is about narrowing candidates or about expecting compression. None of them is about hearing the syllable more accurately.
- 01Two syllable words, in isolationThe second syllable is what disambiguates the first
- 02One clip, twenty timesUntil the compressed tones stop being surprising
- 03The transcript, after the failureTo find the two or three places you and it disagree
- 04A familiar subject at full speedWhere prediction has something to run on
- 05An unfamiliar subject, with a faceThe last support before the telephone removes it
The planks are in the order they get useful rather than in order of difficulty. The final bank is genuinely harder than anything before it, and a learner who is comfortable everywhere else and lost on the phone has not regressed.
Two sources, and one of them is unusually generous
Chinese television is subtitled in characters as a matter of course, because the country contains speakers of many varieties, and that convention is a gift to learners that no other major language offers at the same scale. It replaced hunting for a transcript, which for most languages is the hard part. Watch with the subtitles, then again without, and the second pass is where the listening happens.
Du Chinese is the graded half: texts at stated levels with audio, so the same material can be read and heard at a density where almost everything is known. It replaced the recorded textbook dialogue, which is written Chinese performed slowly and teaches a register nobody speaks. Free tier with a paid subscription above it. The point of a graded audio text is not the content, it is that the vocabulary is already yours, so the only new thing being practised is the hearing.
The mistake: listening once, to a lot of things
The instinct is to consume volume, because volume is what everybody recommends and because a fresh clip feels like progress. A single pass over new material rarely reaches the point where the compressed tones become expected, so twenty clips heard once each teach less than one clip heard twenty times.
The version that works is narrow and repetitive. Take thirty seconds, listen until it is comfortable, then take the transcript and find the two or three places where what you heard and what is written disagree. Those places are the sandhi, the neutral tones and the compressed syllables, and they are the entire lesson. Then move to something slightly faster. It is dull, it is the thing that moves this skill, and it is why a learner with two hundred hours of background listening can still be behind one with twenty hours of deliberate listening.
What listening rests on
The syllable being heard was recorded in a Roman spelling whose values are frequently misread, and a syllable stored wrong is a syllable that will never be recognised, which is the Roman spelling, and where it misleads English readers. The tone arriving on it behaves differently in a sentence than in a drill, which is four of them, and what happens when you get one wrong. The words being selected are mostly two syllable compounds, and the second syllable is what disambiguates the first, which is two character words, and guessing the meaning of a new one. And where the hours come from in an ordinary week is switching your phone, your music and your commute.