Spaced Repetition: Why It Works and How to Use It

• Spacing beats massing at equal total effort. Established since 1885 and replicated hundreds of times.
• Optimal review gaps are roughly 10–20% of how long you need to remember.
• Spaced repetition combines the spacing effect with the testing effect — retrieval, not re-reading.
• The difficulty is the point. Bjork's "desirable difficulties" make practice feel worse and retention better.
• It works on discrete retrievable units. It is not a substitute for understanding.

The forgetting curve

In 1885, Hermann Ebbinghaus ran an experiment on the only subject he could fully control: himself. He memorised lists of nonsense syllables — deliberately meaningless, so prior knowledge could not help — then measured how much he retained at intervals.

The result was the forgetting curve: memory decays steeply at first and then flattens. Most of what you lose, you lose quickly. Something you know perfectly at the end of a study session may be substantially gone within a day.

Ebbinghaus found the second thing too. When he relearned a list, forgetting slowed. And when he distributed his repetitions across days rather than concentrating them, he needed fewer total repetitions to reach the same retention. He had identified both the problem and the solution in the same set of experiments, and the field spent the following century confirming he was right.

The modern confirmation is substantial. A 2006 meta-analysis by Cepeda and colleagues in Psychological Bulletin pooled over 250 studies and found a consistent advantage for distributed over massed practice. And in the major review of study techniques by Dunlosky and colleagues (2013), distributed practice was one of only two techniques out of ten to receive a high utility rating — the other being practice testing.

Two effects, combined

"Spaced repetition" as practised today is really two findings stacked.

The spacing effect concerns when you review: spread out beats bunched up.

The testing effect concerns how you review: retrieving information from memory strengthens it far more than reading it again. Trying to recall a fact and succeeding is a different cognitive act from seeing the fact, and only the first one meaningfully consolidates memory.

Spaced repetition times an act of retrieval for the moment just before you would have forgotten. That timing is what makes it efficient: review too early and the retrieval is trivial, adding little; review too late and you have already forgotten and must relearn from scratch.

Two mechanisms are usually offered for why this works, and they are not mutually exclusive. Study-phase retrieval holds that a spaced review requires you to reconstruct the memory, and that reconstruction is itself the strengthening act. Encoding variability holds that reviews in different contexts — different rooms, moods, times of day — attach the memory to more retrieval cues, giving you more routes back to it later.

How far apart

The most useful practical finding is that the optimal gap scales with how long you need to remember.

In work published in Psychological Science in 2008, Cepeda and colleagues tested a range of gaps against a range of retention intervals and found the best gap was roughly 10 to 20% of the target retention interval:

You need to remember for	Review roughly every
1 week	1 day
1 month	3–7 days
6 months	3 weeks
1 year	1–2 months

The direct consequence for anyone studying for an exam: if the exam is in a month, cramming the night before is the worst available schedule, and it is the one nearly everyone uses.

There is a second consequence that is easy to miss. The optimum is a broad plateau rather than a spike, and the penalty for reviewing too late is much smaller than the penalty for reviewing too early. If you are unsure, wait longer. A slightly over-long gap costs little; a bunched-up schedule costs most of the benefit.

Expanding intervals and the software

In practice, gaps should grow as an item becomes better learned. A card you have recalled correctly five times does not need the same attention as one you met yesterday.

Sebastian Leitner formalised the low-tech version in the 1970s: a box with several compartments. New cards start in the first. Get one right and it moves to the next compartment, which is reviewed less often. Get one wrong and it goes back to the first. The system self-sorts, so effort concentrates where it is needed.

Software automates this per item. Piotr Woźniak's SM-2 algorithm, developed for SuperMemo in the 1980s, adjusts each card's interval based on how difficult you report the recall being, and it or its descendants underpin most modern tools. Anki, the most widely used free implementation, has more recently added FSRS, a scheduler fitted to large volumes of real review data that predicts recall probability more accurately and typically reduces workload for the same retention.

The practical case for software is not sophistication. It is that scheduling several hundred items individually by hand is the part that becomes unmanageable, and it is the part a computer does perfectly.

Why it feels bad

Spaced repetition is uncomfortable, and people abandon it for that reason. Understanding why the discomfort exists makes it much easier to tolerate.

Robert Bjork's research programme introduced the idea of desirable difficulties: conditions that make practice harder and slower while improving long-term retention. Spacing is one. Retrieval practice is another. Interleaving different topics rather than blocking them is a third.

What unites them is a gap between performance during learning and learning itself. Massed practice produces rapid, visible improvement within a session, which feels like progress. Spaced practice produces halting, error-prone sessions that feel like failure. The first is largely temporary; the second is durable.

This creates a genuine metacognitive trap: the method that feels most effective is the least effective, and your own judgement will consistently mislead you. Studies asking learners to predict which condition helped them most find they reliably choose the massed condition — the one they performed worse in.

The point where you struggle to recall something and then get it is not the method failing. It is the method working.

What it is good for, and what it is not

Spaced repetition works on material that decomposes into discrete retrievable units:

• Vocabulary in another language
• Anatomical and medical terminology
• Formulae, constants, conversions
• Legal definitions and case names
• Historical dates, chronologies
• Programming syntax and API signatures
• Musical intervals, chord shapes, note names

It is much weaker as a route to understanding. You cannot make sense of a difficult proof by drilling flashcards of its steps, and a deck of definitions about a subject you do not understand produces a person who can recite definitions.

The productive sequence is: understand first, then retain. Work through the material properly — explain it, apply it, solve problems with it — and then use spaced repetition to keep the components available so your understanding does not quietly decay. Cards written after you understand something also tend to be far better cards, because you know which distinctions matter.

Making it survive a real week

Most people who try spaced repetition quit within a month. The failure modes are predictable.

Too many cards. Reviews compound: every card you add today creates reviews for years. A deck that grows unchecked produces a daily queue that becomes a second job, and then you stop. Add fewer cards than feels right. Ten a day is 3,650 a year.

Cards that are too big. "Describe the causes of the First World War" is not a flashcard; it is an essay prompt. One card, one fact, one retrieval. Break large things into small ones.

Cards you do not understand. Memorising a sentence you cannot explain produces a person who can recite a sentence. If a card keeps failing, the problem is usually comprehension, not memory — go back to the source.

Skipping days. The queue does not forgive: three days off produces a triple-sized backlog that is intimidating enough to trigger quitting. Better to do a shorter session every day than a long one sometimes. If you do fall behind, do not try to clear the backlog in one sitting — cap the session and let it drain over a week.

Passive review. Flipping a card and thinking "yes, I knew that" is re-reading with extra steps. Say the answer out loud, or write it, before you turn the card. The retrieval is the whole mechanism, and skipping it wastes the session.

Further reading: Dunlosky et al.'s Improving Students' Learning With Effective Learning Techniques (2013) surveys the evidence for ten study methods. Robert Bjork's Learning and Forgetting Lab publishes accessible summaries of the desirable-difficulties research. Anki is free on desktop and Android.