What is spaced repetition, and why does it work?

8 min read·Updated

Spaced repetition is a study method that shows you each fact at widening intervals — a day, then three days, then a week — timed to land just before you would have forgotten it. Each successful recall at the edge of forgetting strengthens the memory more than an easy one would, so the intervals can keep growing. It is the most efficient known method for retaining large sets of discrete facts, such as vocabulary.

Key points

  • Reviewing just before you forget produces more durable memory than reviewing while the fact is still fresh.
  • Retrieval — pulling the answer from memory — does the work. Rereading does very little.
  • Difficulty is the point: a review that feels slightly hard is a review that is working.
  • The schedule only helps if you keep to it, which is why daily queue size matters more than daily study time.

The forgetting curve

In the 1880s Hermann Ebbinghaus memorised long lists of nonsense syllables and tested himself at intervals to chart how quickly they faded. The result — a steep initial drop that levels off — became known as the forgetting curve, and it has held up across more than a century of replication.

The practical consequence is uncomfortable. Most of what you learn in a study session is gone within days unless something interrupts the decay. Rereading your notes feels like it interrupts it, but mostly it produces familiarity rather than memory: you recognise the material and mistake recognition for recall.

Ebbinghaus also found the fix in the same experiments. Each time he relearned a list, it decayed more slowly than before. Repeated exposure flattens the curve — and the timing of those exposures turns out to matter enormously.

The spacing effect

Given a fixed amount of study time, spreading it across several sessions produces substantially better long-term retention than spending it all at once. This is the spacing effect, and it is one of the most robust findings in experimental psychology — reproduced across ages, subjects and decades.

The counterintuitive part is that spacing makes practice feel worse. Cramming produces high performance during the session, which feels like learning. Spaced practice produces more errors in the moment and far better retention weeks later. Learners consistently rate the method that works as the method that is not working.

Why retrieval, not review

Spacing pairs with a second effect: testing yourself is dramatically more effective than restudying. The act of retrieving a memory strengthens it, and a retrieval that takes effort strengthens it more than an easy one.

This is why a flashcard has two sides and why you should not flip it early. The two or three seconds of struggle before you turn the card over is not wasted time — it is the mechanism. Flipping immediately converts an effective retrieval exercise into ineffective rereading.

What a scheduling algorithm actually does

Put the two effects together and you get the design brief for a spaced repetition system: test the learner by retrieval, and time each test to arrive just before the memory would have failed.

Nobody can predict that moment exactly, so the algorithms approximate it. They track how well you have done on each individual card and adjust that card's interval accordingly. Get it right easily and the next gap grows a lot. Struggle and it grows a little. Fail and it collapses back to minutes.

The practical effect is that your daily queue fills with the material you are closest to losing, and everything you know solidly gets out of the way. A deck of two thousand mature cards might surface thirty reviews a day rather than two thousand.

Running it without burning out

The failure mode of spaced repetition is not the algorithm. It is the day you open a backlog of four hundred due cards, close it again, and never return.

Two habits prevent that. First, cap how many new cards you add per day — twenty is a common, sustainable figure, and every new card you add today is a review obligation for years. Second, review daily rather than in weekly blocks, because the schedule assumes you show up roughly when it says.

The other common mistake is bad cards. A card testing five facts at once will fail repeatedly, drag its ease down, and reappear constantly. One fact per card, phrased so there is exactly one correct answer, is worth more than any amount of algorithm tuning.

  • One fact per card. Split anything that has two answers.
  • Cap new cards per day; the backlog compounds.
  • Review daily, even a short session.
  • Grade honestly — inflating grades only pushes the reckoning further out.
  • Delete cards you no longer care about instead of failing them forever.

Where it fits in language learning

Spaced repetition is excellent at discrete facts: a word and its meaning, a character and its reading, a verb and its irregular past form. It is not a substitute for reading, listening or speaking, and a deck alone will not make anyone fluent.

The realistic framing is that it removes vocabulary as a bottleneck so the rest of your practice can work. Reading a novel is far more useful when you know most of the words, and conversation stops being a stalling exercise when the nouns arrive on time.

Frequently asked questions

How long should a daily spaced repetition session take?
For most learners with a controlled number of new cards, ten to twenty minutes a day is enough. Session length is driven mainly by how many new cards you introduce, since each one generates reviews for years afterwards.
Is spaced repetition better than just reading a lot?
They do different jobs. Extensive reading builds intuition, collocation and comfort; spaced repetition guarantees that specific items stick. Most successful learners do both, and use the reading to decide what deserves a card.
What happens if I miss a week?
Cards accumulate and come back overdue. Nothing is lost — work through the backlog in small daily batches rather than all at once, and pause new cards until the queue is clear.
Does spaced repetition work for grammar?
Partially. It works well for discrete grammatical facts — an irregular form, a case ending, which preposition a verb takes. It works poorly for patterns you need to produce fluently, which want drilling and output practice instead.