How to Remember What You Learn: The Methods That Survive Testing
The memory advice that circulates online is mostly a mixture of one nineteenth-century experiment, one Renaissance technique and a great deal of invention. A smaller set of findings has been replicated for decades — and they are less comfortable than the tricks.
pixelroad · First published 10 March 2021 · Updated 9 September 2026
Almost everything popularly known about memory descends from two sources: a German psychologist who memorised nonsense syllables alone in the 1880s, and a Renaissance tradition of imagining objects placed inside imaginary buildings. Both are genuinely interesting. Neither is the reason you have forgotten the German dative.
The findings that have survived a century of testing are narrower, duller and considerably more demanding. They share an awkward property: the methods that work best are the ones that feel least productive while you are doing them.
The finding that matters most: spacing
If you learn something twice, the gap between the two encounters affects how long you retain it. A large meta-analysis by Cepeda and colleagues, drawing together decades of studies, found that distributed practice reliably outperformed massed practice, and that longer gaps favour longer retention. Reviewing on Monday and again three days later beats reviewing twice on Monday, for the same total effort.
This is the single most robust result in the field, and it is the one most systematically ignored, because massed study produces the pleasant sensation of fluency and spaced study produces the unpleasant sensation of having forgotten.
The second: retrieval, not review
Roediger and Karpicke's work established what is now called the testing effect: attempting to recall material strengthens memory more than re-reading it, even when the attempt fails. Reading a vocabulary list four times feels like learning. Covering it and trying to produce the words feels like failing. The second is the one that works.
Applied to a language, this means the flashcard must be answered before it is turned over, the verb table must be written from memory before it is checked, and the phrase must be said aloud before the recording is replayed.
The third: interleaving
Studying one thing until it is finished, then moving to the next, produces better performance during practice and worse performance later. Rohrer and Taylor's work on mathematics practice found that mixing problem types — so the learner must first work out which method applies — improved later test performance, despite feeling clumsier at the time.
For a language, interleaving means mixing tenses in one exercise rather than drilling the perfect for forty minutes. Drilling one tense teaches you to conjugate. Mixing them teaches you to choose, which is the thing you actually have to do in a conversation.
Why the good methods feel bad
Robert Bjork gave this pattern a name: desirable difficulties. Conditions that slow learning down and increase errors during practice frequently improve long-term retention. Conditions that make practice feel smooth — rereading, massing, blocking by topic — frequently produce fluency that evaporates within a week.
The practical consequence is uncomfortable. Your sense of how well a study session went is a poor guide to how much of it you will keep, and it is systematically biased towards the least effective methods.
Your sense of how well a study session went is a poor guide to how much of it you will keep — and it is biased in favour of the methods that work least well.
Learning styles — the idea that people learn better when material matches a visual, auditory or kinaesthetic preference — has been tested repeatedly and is not supported. The claim that we use ten per cent of our brains is simply false. The forgetting curve is a real observation from Ebbinghaus's experiments on himself, later replicated in shape by Murre and Dros in another single-subject study; it is not a precise schedule for when you will forget a French word. Sleep does appear to matter for consolidation, but the effects are conditional and the literature is more contested than the headlines suggest, and meditation research supports modest attention and working-memory effects rather than a general memory improvement. No app can guarantee an optimal algorithm for your brain, and any product that claims to has left the evidence behind.
Where memory palaces genuinely help
The method of loci is not folklore: placing items along a familiar imagined route measurably improves recall of unrelated lists, and competitive memorisers use it because it works for exactly that task. What it has not been shown to do is build the automatic, fast, contextual access to vocabulary that speaking a language requires.
Which suggests a limited but real use: a palace is a reasonable tool for the small set of things that must be held in a fixed order and refuse to stick — the six Russian cases, the German prepositions that take the dative, an irregular list you have failed to memorise four times. It is a poor tool for two thousand words.
A method you can actually run
Nakata and colleagues tested spaced retrieval in ordinary classroom vocabulary learning rather than in a laboratory and found that the pattern held up under real conditions. That is the version worth copying: not an elaborate system, but a small number of habits arranged so that the difficulty falls in the right place.
Answer before you check. Leave a gap of days rather than minutes. Mix the categories in a single session. Return to the items you failed, and stop returning to the ones you have not failed in a month. Accept that a session which felt hard was probably the one that worked.
The part memory work cannot reach
All of this concerns retention of things you have correctly learned. None of it detects a structure you have learned wrongly. A learner can space, retrieve and interleave an error for a year with perfect discipline and simply install it more firmly — which is the one failure mode that no amount of private study can catch.
Spacing and retrieval are things you can do alone. Being caught in a mistake you did not know you were making is not.
See our language courses