Here is an experiment that takes ninety seconds.
Pick something you know well. Your job, a subject you studied, the plot of a film you love. Now explain it out loud, to nobody, as though the person listening has never heard of it.
Most people get about forty seconds in and discover a hole.
Recognition is not recall
The reason is a distinction psychologists have been circling for decades: recognising something and producing it are different mental operations, and they can come apart dramatically.
Recognition is what happens when you reread a chapter and everything looks familiar. It feels like knowledge. It generates confidence. And it is a poor predictor of whether you can produce the same information when nothing is in front of you.
Production is harder because you have to generate the material rather than confirm it. There is no page to lean on, no multiple-choice option that looks more right than the others, no diagram doing the structural work. If a piece is missing you find out immediately, because the sentence stops.
The uncomfortable version: the study methods that feel most productive — rereading, highlighting, watching a lecture again — are the ones that train recognition most efficiently and recall least.
This has a name in the research literature. Speaking material aloud tends to make it more memorable than reading it silently, an effect robust enough to have been replicated in a lot of settings. Testing yourself on material beats reviewing it, even when the review takes longer. Both findings point the same direction, and both are widely ignored in practice, because both feel worse while you are doing them.
Why it feels harder
Difficulty is the mechanism, not a side effect.
When you say something out loud you are doing several things at once: retrieving the content, sequencing it, monitoring whether it makes sense, and listening to yourself in real time. That last one is the part that has no equivalent in silent study. You hear the answer arrive, and you hear the places where it does not.
This is why people find their fourth attempt at an answer dramatically better than their first, and why they are usually surprised by it. Nothing new was learned between attempts. The material was reorganised by the act of producing it.
Programmers have known a version of this for years. Rubber duck debugging — the practice of explaining your broken code, line by line, to an inanimate object on your desk — works often enough to have become a standard technique. The duck contributes nothing. The explaining is the entire intervention.
What this means if you are preparing for something
Three practical consequences, in rough order of how much they matter.
Producing beats reviewing. If you have an hour, spending forty minutes reviewing and twenty minutes producing will generally serve you worse than the reverse. This is true for exam material, interview answers, presentations, and languages — anything where you will eventually have to generate rather than recognise.
The first attempt is supposed to be bad. People stop after one bad attempt because they take it as evidence they are not ready. It is evidence that producing is a separate skill from knowing, which is exactly the thing you sat down to fix.
You need something that does not help you. A friend fills your silences and finishes your sentences, because that is what considerate people do in conversation. It makes the practice pleasant and removes the difficulty that was doing the work.
Where this does not apply
Worth being honest about the limits.
If you do not yet know the material, production practice is not the answer — you cannot produce what you have not learned, and grinding at it is demoralising rather than useful. Build the raw material first, then switch.
And the effect is about retrieval, not comprehension. Saying a wrong thing fluently makes it more memorable, which is not the same as making it correct. Production practice needs something checking the content, whether that is a textbook, a colleague, or something that asks you why after every answer.
The general principle is unglamorous and holds up well: whatever you will eventually have to do, practise that thing. If the exam is spoken, speak. If the interview is a conversation, have conversations. If you will have to explain it, explain it — to a person, to a duck, or to something on the phone that will let you be bad at it first.