Research · Memory

The forgetting curve, and how
spaced reviews flatten it

The forgetting curve is the line that shows how memory of something learned once fades over time: fast at first, then more and more slowly. Hermann Ebbinghaus measured it on himself in the 1880s, and it has been reproduced since. Here is what he did, what his numbers mean (and what they do not), and how reviews at growing gaps change the curve.

Start your notebookTry the calculator

By , who makes Lexpad. Checked against the app on 25 October 2026.

What Ebbinghaus did

Ebbinghaus wanted to measure memory without meaning getting in the way, so he learned lists of nonsense syllables, 13 to a list, until he could recite them twice without error. He was his own only subject. After a set time he learned the same lists again and measured how much less time it took. That difference, the savings, was his measure of what was still in memory: even a list he could not recite was partly kept if it came back faster (Ebbinghaus, Memory, 1885, chapter 7).

The curve, in his numbers

His savings fell quickly in the first hour, more slowly over the first day, and very slowly after that. A month later, he still saved about a fifth of the original work.

Ebbinghaus’s savings in 1885: the share of the first learning time he saved when he relearned the lists after each interval (Memory, chapter 7).
Time since learningSavings
20 minutes58.2%
1 hour44.2%
8.8 hours35.8%
1 day33.7%
2 days27.8%
6 days25.4%
31 days21.1%

These numbers are savings, not the share of words remembered. A popular claim that people forget “70% within 24 hours” or “90% within a week” is often credited to Ebbinghaus; his data are about relearning time for nonsense syllables, so they do not say that about recall of meaningful material.

The 2015 replication

In 2015 Jaap Murre and Joeri Dros repeated the experiment as closely as they could: one subject (Dros himself), about 70 hours of learning and relearning lists of nonsense syllables, at the same intervals from 20 minutes to 31 days. The curve came out very like Ebbinghaus’s. Fitting their data with his and an earlier German replication, they found signs of a small upward jump at one day, which may be linked to sleep; they say that still needs testing (Murre and Dros, PLOS ONE 10(7), 2015). After 31 days their subject kept less than Ebbinghaus had; the authors discuss why.

What reviews do to the curve

Each time you bring a memory back, it fades more slowly afterwards. This is the idea behind spaced repetition: review a word just as it starts to slip, then wait longer before the next review. The chart shows it with a modern memory model.

Modelled chance of recalling a word over 90 daysTwo lines. Studied once on day 0: the chance falls to about 57% by day 90. Reviewed on days 1, 4, 11, 32 and 77 and answered right each time: the chance stays above 94% and ends near 99%.50%60%70%80%90%100%0153045607590Days after you first learn the word
Solid line, dots: reviewed on Lexpad’s box days (1, 4, 11, 32 and 77) and answered right each time. Dashed line: studied once. A model, not learner data: FSRS-6 with its standard parameters, worked out when this page was built.
The same model in numbers: the chance of recalling the word on that day, before any review that day.
DayStudied onceReviewed on the box days
195%95%
486%95%
1177%96%
3266%94%
6060%97%
9057%99%

How it was drawn: the model is FSRS-6 with its standard parameters (FSRS, The Algorithm), the same one Lexpad’s Smart schedule and the spaced repetition calculator use. It estimates the chance of recalling a word on a flashcard, for a word answered right on day 0 and at each review. The review days are Lexpad’s five boxes: today, then 1, 3, 7 and 21 days later, then 45 days once mastered. Real words differ: some are easy, some keep slipping, and a miss starts the word again from box 1.

What the curve shows, and its limits

  • Nonsense syllables are hard on purpose. Words with meaning, a sentence and a reason to know them are usually remembered better.
  • One person, one kind of material. Ebbinghaus and the replication each tested a single subject. The shape has held up widely; the exact numbers belong to their experiments.
  • It is a family of curves. How fast you forget depends on the material, how well you learned it, and how often it has come back. Spaced repetition models, such as FSRS, estimate a curve for each card from your answers.

Using it for vocabulary

  • Review the next day. The steepest part of the curve is the first day.
  • Then let the gaps grow: a few days, a week, three weeks. A word that comes back right just as it was fading lasts longer the next time.
  • Test, do not reread. Recalling a word slows the forgetting more than looking at it again (active recall for vocabulary).
  • Keep the sentence. Meaning and context are what nonsense syllables lack; they help a word stay.

The schedules in detail: spaced repetition for vocabulary, the Leitner system and Leitner boxes or FSRS.

How Lexpad uses it

Made by the author of this page

Each word comes back before it fades

  • Lexpad brings each word back on the box days in the chart: today, then after 1, 3, 7 and 21 days while you know it. A mastered word then waits 45, 90, 180 and 365 days; a miss sends it back to box 1.
  • Or choose Smart (Beta) for a notebook: it uses FSRS-6 to work out each word’s own gap from your answers, and brings it back when your chance of recalling it falls to the target you set (85, 90 or 95%).
Open in LexpadAll downloads

Free account, no card. Apps: Mac · Windows · Android · iPhone

Sources, read 25 October 2026

Good to know

Questions

What is the Ebbinghaus forgetting curve?

A curve showing how memory of something learned once fades over time: steeply in the first hour and day, then more slowly. Ebbinghaus measured it on himself in 1885 by relearning lists of nonsense syllables.

Do we forget 70% of what we learn in 24 hours?

That figure is not from Ebbinghaus’s data. His numbers measure how much time relearning saved (about 34% after one day) for nonsense syllables, which is a different thing from the share of meaningful material you can recall.

Has the forgetting curve been replicated?

Yes. Murre and Dros repeated the experiment in 2015 (PLOS ONE) with one subject and found a very similar curve.

How do you beat the forgetting curve?

Review what you learned before it fades, at growing gaps: the next day, a few days later, a week, three weeks. Test yourself rather than rereading.

How is the forgetting curve used in spaced repetition apps?

Schedulers such as FSRS estimate a forgetting curve for each card from your answers, and plan the next review for when your chance of recalling it falls to a target.

Lexpad

Keep the words you meet.

Save the words you meet with their sentence; AI writes the card and spaced repetition brings each one back. Free to start, on the web, Android, Mac and Windows.

Every new account gets Pro free for 7 days, no card needed.

Open in LexpadAll downloads

Free account, no card. Apps: Mac · Windows · Android · iPhone