Retrieval practice

Why does trying to recall an answer improve learning more than rereading it?

Almost everyone studies by rereading. Underneath sits an assumption too obvious to state: that memory is a bucket, learning is pouring in, and testing is checking the level. On that picture, time spent testing is time not spent learning — measuring, not filling.

Notice what the picture predicts: more exposure should always beat less. Yet the finding that survives repeated replication runs the other way. After the same total time, students who spent part of it trying to recall the material remember substantially more a week later than students who spent all of it rereading. If the bucket picture were right, that should be impossible.

Which part? Consider what happens when you reread: the words are in front of you, so nothing has to be found. Now close the book and try to produce the answer. Something has to be searched for. Is that search a neutral inspection, or an event in its own right?

Push on it. You look up a phone number and copy it down, versus dialling it from memory. Which leaves you more likely to have it tomorrow? Most people know from experience that it is the second — even though the first involved staring at the digits longer. So pulling something out appears to change what is there, rather than merely reporting on it.

Why might that be? Two mechanisms are proposed, and the mechanism is still argued over even though the effect is not. One is that a successful search does work: to find the memory you generate cues, connections, partial routes, and those routes are left behind as extra ways in. The other is Bjork's distinction between how well something is stored and how easily it can currently be reached; retrieval strengthens reachability specifically, and most when the reaching was hard. That is the "desirable difficulty" idea, and it fits a suspicious pattern — retrieval helps more when effortful and less when the answer is already on the tip of the tongue.

Now the trap, where most of the practical damage happens. Which method feels more productive while you are doing it? Rereading, obviously — smooth, familiar, error-free. Struggling to recall feels like failure, because it produces visible blanks and wrong answers. So the better technique feels worse, and students choosing by feel choose wrongly, mistaking the fluency of familiar text for evidence of retrievability.

Two honest limits. Retrieval without feedback is much weaker and can consolidate a confidently wrong answer, so checking afterwards is not optional. And most of the strongest demonstrations use word pairs, vocabulary, and short passages tested by recall. The extension to complex, inference-heavy material — and to problems different from the ones practised — is real but smaller, more variable, and genuinely less settled. Retrieval practice reliably makes you better at retrieving what you practised retrieving; how far that carries into flexible understanding is the part still being worked out.

Two refinements. First, "testing" here means any attempt to produce the answer from memory — self-quizzing, covering the page and reciting, working a problem without the worked example, explaining it aloud. The graded exam is one instance of the thing, not the thing itself, and most of the benefit needs no grading at all. Second, the effect is about the attempt, not the score: an unsuccessful retrieval followed by the correct answer still generally beats simply being shown the answer, which is the strongest evidence against reading it as mere extra exposure.

One caution: repeatedly retrieving part of a set can temporarily make closely related, unpractised items harder to recall — a real laboratory phenomenon whose classroom importance is debated. Practising narrowly and assuming the rest came along for free is the failure mode to watch.

• Why spacing the retrievals out matters as much as doing them, and how the two effects combine.
• What changes when retrieval practice is used for skills and procedures rather than facts.
• Whether pre-testing — guessing before you have been taught — produces the same benefit.

Every term the collection defines is gathered in the glossary.