Listening practice that actually works (and why subtitles ruin it)

Here is a test worth running on yourself. Put on something in your target language with subtitles in your own language, watch for five minutes, then turn the subtitles off and watch five more.

If comprehension falls off a cliff, you were not listening. You were reading, with audio playing in the background.

Why subtitles quietly cancel the exercise

Reading is faster than listening and requires less effort, so given both at once your brain takes the cheap route every time. You do not choose this and you cannot override it by trying harder. The text wins.

The consequence is that a hundred hours of subtitled viewing can leave your listening exactly where it started. People conclude they have "bad ears". They do not. They have never once been made to decode sound without a safety net.

What listening-first means in practice

The fix is structural: hear the line before you are allowed to see it.

That order matters more than it sounds. When the text is hidden, your brain has to commit — it produces a guess, and a guess is a prediction that can be right or wrong. Revealing the text a second later either confirms or corrects that prediction, and correction against a committed guess is roughly where all learning happens.

Flip the order and nothing happens. Text first, audio second, is just reading aloud.

This is what the Listen first setting in Amiko does: the character's line arrives blurred, you play it, you decide what you heard, and then you uncover it. It costs one extra tap and changes the exercise completely.

Slow it down, but only once

Native pace collapses words together. What do you want to do becomes whaddaya wanna do, and the words you are listening for are physically not there as separate sounds.

Slowed playback separates them again so you can hear the components. But do not stay there — the goal is normal speed, and slow audio practised forever trains you for a version of the language nobody speaks. A useful rhythm: normal, slow once, then normal again. The second normal pass is the one that counts, and you will often hear the line clearly on it after failing twice.

Amiko's audio plays a touch slower than a native would speak on the charts, where clarity is the whole point, and at natural speed in conversation, where realism is.

Ten minutes beats an hour

Listening at the edge of your ability is genuinely tiring, and tired listening is passive listening — you drift, the audio continues, and the session becomes background noise.

Ten concentrated minutes where you are guessing, checking and being wrong will move you further than an hour of a podcast you have stopped following. If you can still feel the effort, it is working. If it has become comfortable, either the material is too easy or you have stopped paying attention.

Use material where you already know the situation

Context does an enormous amount of the work. If you know a scene is someone ordering coffee, half the possible sentences are eliminated before anyone speaks, and your guesses get dramatically better.

This is why conversation is a better listening exercise than a podcast when you are starting out: you know what you just said, so you can predict what is coming back. A roleplay conversation in a shop or a café gives you the situational scaffolding that unstructured audio does not.

The thing to listen for is not words

Beginners try to catch every word and lose the sentence while stuck on the third one. Better listeners catch the shape first: who is doing what, is this a question, is it past or future, is the tone friendly or annoyed.

Get the shape and the words fill in. Chase the words and you get neither. Practically: after each line, before revealing anything, answer one question — "what did they want?" That single habit reorganises how you listen.

A ten-minute routine

Play the line with the text hidden. Guess. Reveal, and see how close you were. Where you were wrong, play it once slowly to find the sound you missed, then once more at normal speed. Move on.

Five or six lines a day, done that way, is a real listening practice. It is unglamorous, it takes ten minutes, and it is the only version of this that has ever worked reliably.