Why You Can Read a Language and Still Not Hear It

You can read a newspaper column in French with a dictionary open twice. Then someone at the next table says nine words and you catch two of them, one of which turns out to be wrong.

The usual explanation is that your vocabulary is too small. It almost never is. You knew every word in that sentence. You could not find them, because they did not arrive as words.

Speech does not contain spaces

Print gives you the segmentation for free. Speech does not. The signal is continuous, and the listener has to cut it into units while it is still moving — at conversational speed, a few syllables per second, with no opportunity to go back.

Worse, the units change shape. Connected speech reduces, contracts and blurs: French runs words together and swallows vowels, English turns going to into a single mumble, Spanish loses consonants at the end of syllables. The word you learned in its careful dictionary form is not the word that reaches your ear. You are hunting for something that is not there.

The applied linguist John Field, who has spent his career on second-language listening, argues precisely this: that listening instruction has too often tested comprehension instead of teaching the decoding that comprehension depends on. Playing a recording and asking questions about it measures the problem. It does not treat it.

Why more exposure alone stalls

Background listening feels productive and is mostly not. A podcast at breakfast that you follow at fifty per cent will still be fifty per cent in six months, because nothing in the routine forces the unresolved fragments to resolve. The brain is content to guess from context, and guessing from context is a skill you already have in your own language.

What moves the needle is deliberately narrow work on the parts you missed — and, usefully, the research supports one specific tool.

Captions in the same language

A meta-analysis by Montero Pérez and colleagues, covering eighteen studies of captioned video, found that captions benefit both listening comprehension and vocabulary learning. The mechanism matters: same-language captions let you see the boundaries you could not hear, so the sound and the segmentation are aligned in the same moment.

Subtitles in your own language do something else entirely. They let you follow the story while your ear disengages, which is a pleasant way to watch a film and a poor way to train.

A routine that repairs decoding

Work short and hard. Two minutes of audio worked closely beats an hour played in the background. Choose material slightly above your level, not far above it.

Listen three times before you look. Once for gist, once for the specific point where you lost the thread, once trying to reconstruct the exact words. Only then read the transcript or turn on same-language captions.

Find the seam. When the transcript reveals what was said, ask why you missed it. Unknown word, or known word in an unrecognised form? The second is far more common, and it is the one that repays attention.

Say it back. Replay the sentence and repeat it at full speed, copying the reductions rather than pronouncing each word neatly. Producing connected speech teaches you to expect it.

Then go wide. Once a fortnight of close work is in place, ordinary listening starts to pay, because you now hear the joins.

What improvement feels like

Not a sudden clearing. It comes as a shortening delay: the sentence lands, and instead of being lost it is understood a beat late. Then half a beat. Then it is simply understood, and you stop noticing the language at all — which is, in the end, the whole point of learning one.