The forgetting curve: what it is, and how to beat it

What the forgetting curve is

The forgetting curve describes how much of something you've learned is still in memory as time passes. Draw it and you get a line that falls steeply in the first hours and days, then levels off: most of what you'll lose, you lose early, and what survives the first few days fades much more slowly.

It was first measured by Hermann Ebbinghaus, a German psychologist, who published his results in 1885 in “Über das Gedächtnis” (translated as “Memory: A Contribution to Experimental Psychology”). It's still the starting point for every study of memory over time, and the reason flashcard apps schedule reviews the way they do.

How Ebbinghaus measured it

Ebbinghaus was his own subject. To avoid the head start that meaning gives, he learned lists of nonsense syllables, such as “dax” or “bok”, reading each list aloud until he could recite it twice without a mistake. Later, after a set delay, he learned the same list again and timed how long it took.

His measure was savings: how much less time the relearning took, as a share of the original learning. If a list took ten minutes to learn and four to relearn, 60% of the work had been saved, a sign that much of the memory was still there even if he couldn't recite the list. It's a sensitive measure: it catches memory you can't consciously recall but that still makes relearning faster.

What he found

For lists of 13 syllables, the work saved on relearning fell like this:

Two things stand out. More than half of the loss happened within the first hour. And after the first day, the curve flattened dramatically: between day one and day 31, savings fell only from about a third to about a fifth. Forgetting is fast, then slow.

A caution about the numbers: they describe one man learning meaningless syllables, and savings isn't the same as the share you could answer on a test. Meaningful material, like your course, is easier to learn and to keep. The shape of the curve, steep then flat, is what carries over.

Has it held up?

Yes. In 2015, Jaap Murre and Joeri Dros replicated the experiment as closely as they could: one participant spent 70 hours learning lists of syllables and relearning them after 20 minutes, 1 hour, 9 hours, 1 day, 2 days or 31 days. The results were similar to Ebbinghaus's.

Their analysis found one interesting wrinkle: the curve isn't perfectly smooth, but seems to rise slightly at the one-day point, after a night's sleep. The authors are careful not to claim sleep is the cause, but it fits a large body of research on sleep and memory, summarised by Susanne Diekelmann and Jan Born in 2010: sleep after learning helps stabilise new memories.

Why we forget

Memories fade without use

A memory that's never retrieved gets harder to reach. The traces are rarely gone completely, which is why relearning is faster than learning, but the path to them weakens.

New learning gets in the way of old, and old of new

Similar material competes. Learn two drug classes with similar names in the same week and they blur together. Psychologists call this interference, and it's why contrast cards (“how does A differ from B?”) are worth making.

The cue is missing

Often the information is stored, but the question you're asked doesn't match how you learned it. Studying an idea from several angles gives you more ways back to it, which is why good flashcards test the same idea in more than one way.

Each review makes the next drop gentler

Here's the part that makes the forgetting curve useful. When you recall something successfully, memory doesn't just go back up to where it was; it becomes more stable, so the next fall is slower. Review again after that longer gap and it slows again. Instead of one curve, you get a series of ever-flatter ones.

Ebbinghaus himself noticed that repetitions spread out over time were more effective than the same repetitions massed together. More than a century of research on the spacing effect has confirmed it. In a 2008 study of more than 1,350 people, Nicholas Cepeda and colleagues found that the best gap before a review grows with how long you need to remember: about 3 days for a test a week away, about 27 days for a test a year away.

Modern schedulers turn this into numbers. FSRS, the algorithm FlashCardz uses, gives each card a stability: the number of days until your chance of recalling it falls to 90%. Each successful review raises the stability, so the gap before the next review grows. For a card you get right every time, the gaps look like this:

That's six reviews to keep a fact for well over a year, compared with dozens of rereads. Harder cards, and cards you forget along the way, get shorter gaps. More in FSRS explained.

Why rereading doesn't flatten it

Not every review is equal. In Henry Roediger and Jeffrey Karpicke's 2006 experiment, students either read a passage four times or read it once and then practised recalling it three times. From a test five minutes later to a test a week later, recall in the reading group fell by 52%, from 83% of the passage to 40%. In the group that had practised recalling, it fell by 14%, from 71% to 61%.

Rereading feels like a review, but it doesn't make memory more stable in the way that recall does. To flatten the curve, the review has to be a test: a flashcard, a question, a blank page. See active recall.

How to beat the forgetting curve

1. Review within a day. The steepest drop is in the first hours, so a short recall session the same evening or the next morning saves the most.
2. Test yourself instead of rereading. Cover the answer, recall it, then check.
3. Let the gaps grow: a day, a few days, a week or two, a month and longer. A spaced repetition app does this per card.
4. Make it meaningful. Ask why each fact is true and connect it to what you already know; connected facts fade more slowly than isolated ones.
5. Sleep. A night's sleep after studying helps lock it in; an all-nighter works against you.

A review schedule you can follow by hand

If you're not using an app, a simple schedule with growing gaps captures most of the benefit. After learning something, review it:

The weakness of a fixed schedule is that every fact gets the same gaps, whether you find it easy or keep forgetting it. That's what the Leitner system partly solves with boxes, and what algorithms such as FSRS solve properly, card by card.

The forgetting curve and your exam date

For an exam, what matters is your memory on one day. Cepeda's results suggest that if the exam is a few weeks away, gaps of several days to a week or two between reviews are about right, and a scheduler should make sure every important card comes up shortly before the exam rather than just after it.

FlashCardz does this when you give a course an exam date: no card is scheduled for after the day before the exam, so everything has been reviewed when it counts. Once the exam is over, the normal schedule returns, so the courses that build on each other, like most of medicine and law, stay with you. See how to study for exams.