Retrieval Practice

Replace re-reading with retrieval. Instead of reviewing notes, ask students to write down everything they remember, then check. The testing is the learning, not the preparation for it.

Retrieval means getting knowledge out of memory rather than putting it in again. In practice it's the difference between a class rereading their notes on the water cycle and a class closing their books and writing down everything they can remember about it. The second one looks less productive and does considerably more.

Five questions per lesson on prior content; nothing in the gradebook. Agarwal found routine quizzing reduces test anxiety rather than causing it. The quiz is not the assessment. It is the learning.

Five questions at the start of every lesson, all of them on content you taught before today, and none of them recorded as a grade. It takes four minutes and it's the single easiest way to get retrieval into a scheme of work without redesigning anything.

A blank page, two minutes, no notes: write everything you remember. The imperfect attempt is the point. What students cannot recall shows precisely where teaching needs to return.

A blank page, two minutes, no notes: write down everything you remember about this topic. No structure imposed, no marks, no right way to lay it out. It's the purest form of free recall available in a classroom.

Quiz students before teaching. Struggling to retrieve primes the brain; students who fail before instruction remember the answer more reliably once taught. Failure is productive.

Quiz the class on material you haven't taught yet. They will get most of it wrong, which is the point. The failed attempt primes them for the teaching that follows.

Students reconstruct key ideas from memory, then open notes and correct in a different colour. Karpicke and Blunt found this outperforms re-reading. The corrections are the data.

Spread retrieval across days and weeks rather than cramming it into one session. Build review points into your medium-term plan. The calendar is part of the pedagogy.

Spread retrieval across days and weeks rather than concentrating it in one session. The same total practice time, distributed differently, produces substantially better retention.

Three questions per lesson: one from yesterday, one from last week, one from last month. These intervals are a starting point; complex topics may need revisiting more frequently.

Three questions at the start of a lesson: one on yesterday's content, one on last week's, one on last month's. It's the simplest possible implementation of spacing and it fits in five minutes.

Kornell and Bjork found mixing topics produces stronger retention than finishing one before moving to the next. Blocked practice feels organised; it produces weaker retention. Discriminating is the mechanism.

Mix question types and topics within a practice session rather than finishing one type before starting the next. Instead of twenty questions on one method, give twenty that switch between three.

Dunlosky's review ranked re-reading among the least effective strategies. Teach students the evidence: study, return after two days, then a week, then a month. Many students do the opposite.

Teach students the schedule itself: study it, come back after two days, then after a week, then after a month. Most students do the opposite, which is to reread everything the night before.

Plan retrieval into your medium-term planning, not just lessons. Bring back earlier content in starters and assessments. When it never reappears, students conclude it no longer matters.

Build earlier content back into starters, homework and assessments across the year, so nothing is finished the moment the unit ends.

Bjork coined desirable difficulties for conditions that slow apparent learning but improve long-term retention. Effortful retrieval is the clearest example. Resist the urge to make it easier.

Choose the version of a task that feels harder in the lesson when the evidence says it retains better. Effortful retrieval is the clearest example, but so are spacing, interleaving and reduced scaffolding.

Retrieval with notes open is reading, not retrieval. Fade support gradually: notes first, then headings only, then nothing. The transition to unaided retrieval is where consolidation happens.

Retrieve the same content in different formats and settings, decontextualising knowledge so it is not tied to one context or format. Harder for students; that is precisely the point.

Retrieve the same knowledge in different formats and settings: written one week, spoken the next, as a diagram the week after, in a different room or a different order.

Slamecka and Graf found generating an answer, even a wrong one, before being told it produces stronger retention than passive exposure. The effort of generation drives encoding, not the accuracy.

Have students produce an answer before they're given it, even when you know the answer will be wrong. Generating beats being told, and generating badly still beats being told.

Do not retrieve immediately after teaching. Wait a day, then quiz. Wait a week, then quiz again. The gap between learning and retrieval is where consolidation happens.

Factual retrieval builds the schema; higher-order retrieval tests whether students can use it. Secure the foundations first, then push beyond them. Retrieval is not a pub quiz.

Retrieval is not only for facts. Once the factual base is secure, retrieve explanations, comparisons, causes and applications too.

Karpicke and Blunt found concept mapping from memory outperformed re-reading for retention. Not just what students know but how ideas connect. Relational knowledge beats isolated facts.

Students draw a concept map of a topic from memory, showing not just the ideas but how they connect. Notes stay closed until the map is done.

The learning happens during preparation, not delivery. Preparing to teach demands complete, organised retrieval. Organising knowledge to explain it is where consolidation occurs.

Have students prepare to teach a topic to someone else. The learning happens in the preparation, whether or not the teaching ever takes place.

Tulving and Thomson showed retrieval depends on available cues. A heading, a keyword, labels removed: each triggers retrieval without doing the work. Builds confidence before unaided retrieval.

Retrieval supported by a cue: a heading, a first letter, a keyword, a diagram with the labels removed. It triggers the search without doing the searching.

Show students their own illusion of knowing. Ask them to rate confidence before revealing the answer. When they were certain and wrong, that is the precise moment to intervene.

Show students the difference between feeling like they know something and actually knowing it. Ask for a confidence rating before the answer is revealed, then let them see the mismatch.

Metcalfe shows instant feedback is essential for confident errors. For general retrieval, allow a processing gap before revealing answers. Speed matters most when confidence is highest.

Reveal the correct answer quickly when a student was confident and wrong. Allow a longer gap when the retrieval was effortful and uncertain.

Dunlosky and Metcalfe showed students are most confident about things most recently read, not things they actually know. Rate confidence after each attempt, then reveal the answer. That gap is the target.

Students rate how confident they're in each answer, then check. Over time the gap between confidence and accuracy narrows, which is a skill in itself.

Zimmerman showed students who monitor their performance make better revision decisions. Log what came back, what did not, and what needs revisiting. It is a map of what to do next.

A running log of what came back easily, what didn't, and what needs revisiting. Kept by the student, not by you.

Dunlosky's review rated ten strategies: re-reading scored low; practice testing scored highest. Ask students how they revise, then show them the data. That conversation is the intervention.

Ask students how they actually revise, then show them what the evidence says about those methods. The conversation is the intervention.