Spaced Repetition

You learned the topic overnight, passed in the morning, walked out pleased. A month later someone asks you about it — and your head is empty. Not "a bit rusty," but genuinely empty, as if you'd never studied it at all.

What's odd is the other side of it: you remember your first teacher's name after twenty years without a single repetition. Some information evaporates in weeks, some holds for decades. The difference isn't importance and isn't volume — it's the schedule on which you return to it.

You lay wood on the embers

Picture a campfire you have to keep going all night.

There are two ways to feed it. The first: dump the whole bundle on at once, while the flame is high. The fire leaps up, it's beautiful for twenty minutes — and then it all burns through in a single flare, and by the middle of the night you're left with ash and not one log in reserve.

The second way: keep the wood beside you and add it one log at a time — not at the height of the flame, but at the moment the fire has almost settled and only glowing embers remain. Every log laid on the embers lifts the fire again, and each time the bed of coals gets thicker and the next pause can be longer. By morning you have a steady fire, and it took less wood than the first way.

Cramming is the whole bundle at once. Spaced repetition is one log on the embers.

And here is the main point, the least obvious one: you add wood when it has almost gone out. Reviewing material you already remember perfectly is throwing logs into a roaring flame: it costs you, it gains you nothing. The benefit appears at exactly the moment when recall comes with difficulty, but still comes.

The formula worth taking away:

You lay wood on the embers, not into the flames.

The forgetting curve

The first person to measure this was Hermann Ebbinghaus in the 1880s — on himself, which makes the story better still. He memorised lists of nonsense syllables, then checked at various intervals how much was left and wrote down the result.

What came out was a curve that has carried his name ever since. Its shape matters more than its exact numbers: the drop is steep at the very start and flatter and flatter afterwards. Most of what's freshly learned is lost in the first hours and days, while what survives a week then lives a great deal longer.

Ebbinghaus noticed a second thing too — the one all of modern practice grew out of: repetition shifts the curve. After each review the material is forgotten more slowly than the time before. The curve keeps flattening. That is precisely why the gaps between reviews can and should be stretched out: first a day, then three, then a week, a month, six months.

The spacing effect

That distributed study beats massed study is one of the most reliably confirmed facts in psychology. In 2006 Nicholas Cepeda and colleagues assembled a meta-analysis of more than two hundred and fifty studies spanning over a century of research, and the result was unambiguous: for the same total time, spaced-out sessions produce markedly better retention than continuous ones.

Then comes the question that matters more in practice than the fact itself: how long should the gap be?

A separate study by the same group (2008) gives a guideline: the optimal gap depends on how long you need the knowledge to last, and it comes to roughly ten to twenty per cent of that span. Need it a week from now — review daily. Need it a year from now — every month or six weeks. The figure is approximate, but the principle holds: the further the horizon, the longer the pauses.

And a word on cramming — it isn't useless, it's narrowly specialised. What you learn overnight really does survive until the morning exam. It just doesn't go much past that morning. If the goal is to pass and forget, cramming honestly works. If the goal is to know, it doesn't work at all.

Why the pause helps

There are several mechanisms and they stack.

Every repetition after a gap is retrieval. While material is still fresh, review collapses into recognition: you simply run your eyes over the familiar. Once the trace has weakened, you have to genuinely pull the knowledge out of memory — and retrieval, as we covered in the third article, strengthens the trace more than rereading. Robert Bjork describes this as a desirable difficulty: the effort of recalling is the mechanism of consolidation.

Each time you recall in a new context. A different day, a different mood, a different room, a different train of thought. A fresh set of hooks attaches to the knowledge — those same labels from the previous article. Material learned in one sitting is tied to one setting; distributed material is tied to a dozen, and you can reach it later by many routes.

Sleep works between sessions. The night shift gets time to file the material away and link it to what's already there. Four hours straight gives the brain not a single night; four one-hour sessions give it three nights of consolidation.

Jeffrey Karpicke and Henry Roediger (2008) showed an additional effect: repeated retrieval after a successful recall keeps strengthening memory, while repeated study of what's already learned adds almost nothing. So it isn't just any log you throw on the embers — it's specifically the attempt to remember.

What it looks like in practice

Running such a schedule in your head is impossible — so it gets automated.

The paper version was devised by Sebastian Leitner in the 1970s: several boxes of cards. A new card goes into the first; recall it correctly and it moves to the next box, which is checked less often; get it wrong and it returns to the first. The material sorts itself by frequency: the hard things come round often, the mastered ones surface rarely.

The digital version — programs like Anki, Mnemosyne and the rest, working on the same principle, except an algorithm sets the interval based on how easily the card came to you. A few rules are worth following:

• One card, one thought. A card with a paragraph of text isn't tested, it's reread.
• The question must demand retrieval, not recognition. Not "is it true that...", but "why...", "how...", "what happens if...".
• Don't peek straight away. Hold for a few seconds — the effort is the work.
• Don't make cards for what you haven't understood. A mechanically memorised phrase you don't follow is a box with no label. First understand, then consolidate.

And the most common place it all falls apart: regularity beats volume. Twenty minutes daily beats three hours once a week by a wide margin. A missed week resets the intervals you've built up — the embers go out and you have to light the fire again.

In practice

First review the same day. The curve falls most steeply in the first hours, so a short return to the material that same evening is disproportionately valuable for the effect and cheap in time.

Then stretch it out: a day → three days → a week → a month. Not a rigid schedule, a guideline. Came back easily — stretch more boldly. Forgot it — shorten the interval and start again from a short one.

Plan hard things in advance, not right before the deadline. You know the date of the exam or interview — count backwards and distribute. This isn't about discipline; it's that otherwise there physically aren't enough nights for consolidation.

Review by retrieving, not by reading. Opening your notes and running your eyes down them isn't review. Review is closing the notes and pulling the contents out of your head, and only then checking.

Check yourself

Close the article and answer in your own words:

1. Why is reviewing well-known material almost useless, while recalling with difficulty is valuable?
2. What is the key feature of the forgetting curve's shape, and what does each repetition do to it?
3. Roughly how do you choose an interval if you need the knowledge to last a year?
4. What are the three reasons a gap between sessions strengthens memory?
5. How does the Leitner system differ from simply rereading a stack of cards in a loop?

In short

• Freshly learned material is lost fastest in the first hours; what survives a week lives long. Each repetition flattens the forgetting curve further.
• Cepeda's meta-analysis: for equal total time, distributed sessions produce substantially better retention than massed ones.
• A guideline for the interval is about 10–20% of the span you need the knowledge for.
• The pause works through three mechanisms: real retrieval instead of recognition, new contextual hooks, and sleep between sessions.
• Review by retrieving, not by reading. Regularity beats volume.
• You lay wood on the embers, not into the flames.