How to actually get something out of shadowing
Shadowing is one of the few practice methods that both interpreter training and language research keep backing up. It works your ear, your sense of rhythm and the muscles in your mouth all at once. But because it asks so much of you, there are more ways to do it wrong than right. This isn't a piece about why it works. It's a routine you can follow today.
Where most people go wrong
- The material is too hard. If you don't understand it, your mouth is copying noise and your head is empty. That's practice at making sounds, not practice at Japanese.
- All reps, no listening back. Half an hour of shadowing and you never once heard yourself. Half an hour of repeating a mistake only sets the mistake.
- Never letting go of the text. While your eyes are reading, your ears aren't working. The text is scaffolding — take it down when the building stands.
Choosing your material
One hard rule: you should follow more than 90% of it without the text. Shadowing trains you to say back what you just heard, so understanding has to be free. If understanding costs you effort, there's no effort left for the shadowing. Which means the material should sit half a level below you: if you're at N3, practise with N4 listening. It isn't a step down, it's what works. Take 30 seconds to a minute at a time, with a transcript, at normal speed — not the slowed-down textbook voice, where the rhythm is fake.
Four stages, one piece of material
- Listen twice with nothing in front of you. Check that both the content and the speed are within reach. If you can't follow it, change the material — don't push through.
- Listen once with the text. Find the words you missed, where the long sounds and small tsu are, and which way the pitch moves. This pass draws the map your mouth will follow.
- Shadow with the text until it runs smoothly. Stay half a second to a second behind, like a shadow. When one line trips you up, work on that line — don't restart the whole piece.
- Drop the text, shadow, and record yourself. This is the real thing. Play your recording next to the original and find the difference with your ears: which beat, which pitch. Then redo that one line.
Stay on the same piece until stage 4 runs smoothly, then move on. One piece done properly beats ten done lazily — everything shadowing gives you is on the climb from rough to smooth. Repeating it after that is a comfort blanket.
How long a day, and when you'll notice
Ten to fifteen minutes, daily. Shadowing is hard enough that quality drops if you go longer — four days of ten minutes beats one session of forty. Don't measure progress by whether you feel fluent. Measure two things you can count: on a fresh piece, stage 3 takes fewer runs than it used to; and when you compare recordings, the faults you can hear get smaller (from "that line was off" to "that beat was too high"). A fussier ear is the progress.
What shadowing can't do for you
One honest note to finish on. Shadowing trains you to sound like someone else's sentence. It does not train you to say your own. Putting a sentence together, pulling up the word you need, going back and forth in a conversation — those only come from actually producing language. Treat shadowing as the whole of speaking and you'll end up a very good tape recorder: lovely pronunciation, and nothing there when it's your turn. It's the foundation, not the house.
Our own plug, clearly marked: the two annoying steps above — finding suitable material with a transcript, and recording yourself to compare — are what manabiu (the desktop app on this site) turns into one path: drag in a video or audio file and you get it line by line (no subtitles? it writes a draft itself), loop any single line to shadow it, and compare any line you pick against a model reading, beat by beat. As for the "say your own" half — that's drilling through to real output, and it's the reason the app exists.