How to do shadowing — and the one thing it will never tell you

What shadowing is, and why it works

Shadowing is speaking along with a recording roughly half a beat behind it. It is not reading aloud from a page. There is no text in front of you: what comes in through your ears goes straight back out of your mouth.

It works because there is no time to store anything. You cannot translate, so you attach to the sound itself — where the speaker breaks, what gets stressed, what gets swallowed. Rhythm arrives in one piece, which is exactly what learning words one at a time never gives you.

It is worth separating what it is good at from what it is not. Rhythm, speed, intonation and linking can shift noticeably within a few weeks. New vocabulary and grammar are a poor fit — shadowing only pays off on sentences you already understand.

Where most of the effort is wasted

Shadowing sentences you do not understand. All you keep is an impression of the sound. If a sentence has more than one unknown word in it, that sentence is too early for you.

Carrying on when you cannot keep up. Mumbling through the hard parts sets the mumble. Clear at a speed you can hold always beats blurred at the original speed.

Never listening back to yourself. This is the common mistake and the expensive one. While you speak, your voice and the recording overlap in your ears, so you cannot hear that the two differ. Only a recording played back, or a score, makes the gap visible.

The real weakness — nothing corrects you while you shadow

Shadowing has no marking scheme. Say an R as an L a hundred times and you have practised the L a hundred times. That is not a flaw in the method; the method simply has no feedback inside it. The feedback has to come from somewhere else.

Tokkly has no screen called shadowing. What it has are two things that fill the hole shadowing leaves. The first is pronunciation scoring in Read aloud: read a sentence out and every phoneme — the smallest unit of sound — gets its own score, with the weak ones explained in terms of where the tongue and lips go. Length is handled separately in Japanese, tone in Chinese (shown as a hint rather than a score), final consonants in Korean. It is unlimited on the free plan.

The second is 🐢 Play slowly in the listening screen. Simply dragging the playback slower smears the sound, so the part you wanted to check is the part you can no longer hear. Here the server re-synthesises the sentence at the slower speed, so it stays clean as it slows down. For a stretch you keep falling off, that is the better tool.

So: shadowing puts the rhythm into your body, and scoring checks the sounds one at a time. Each covers what the other misses.

Step 1 — Pick the clip and settle the meaning (1–2 minutes)

Twenty to forty seconds, or five or six sentences, is plenty. Short material is what lets you go round it many times, and going round many times is where shadowing pays.

Work out the meaning and the structure first, and clear out the unknown words here. Skip this step and everything after it is imitation of noise.

The material can come from anywhere, or you can use sentences straight from Tokkly's listening and read-aloud screens. What matters is not the source but the level — only slightly above where you are now. If you catch less than half of it on a first listen, it is too early.

Step 2 — Cover the text and speak along (4–5 minutes)

The first pass or two with the text in front of you is fine; you need to attach the sounds to the spelling. But do not stay there. From the third pass, cover it. With the text on screen you are practising reading aloud, and as long as your eyes have something to do, your ears take the day off.

Once it is covered, follow the voice half a beat behind on sound alone. When you lose a word, drop it and catch the next sound instead of going back for it. The moment you stop, the rhythm breaks — and the rhythm is the whole substance of this drill.

If you get thrown off three passes in a row, the problem is the speed, not you. Cut the sentences in half, or do one pass slowly and come back.

Step 3 — Measure your own sound, then fix one thing (3–4 minutes)

This is the step almost everyone skips, and the one that turns shadowing into practice. Take one sentence from the clip, say it out loud and get it scored — or record it and play it back against the original.

Pick one or two weak sounds, not all of them. Trying to fix everything changes nothing. Drill only the words containing that sound a few times, then go back and shadow the clip once more.

Sounds get corrected inside words and set inside sentences. Adding that one loop a day changes what shadowing is worth.

Common questions