Guide · Learning science

What is the forgetting curve?

The forgetting curve is a graph of how much of something newly learned you retain as time passes without review. It falls steeply in the first hours and then levels off. The German psychologist Hermann Ebbinghaus measured it on himself and published the results in 1885.

The curve is real and has been replicated. The numbers usually quoted with it are not his. 'You forget 70 percent within 24 hours' misreads both what he measured and what he was memorizing.

How did Ebbinghaus measure forgetting?

Ebbinghaus was his own only subject. To avoid material he already knew, he built about 2,300 nonsense syllables, each a vowel sound between two consonants, and drew them at random into lists. For the forgetting experiment, run in 1879 and 1880, each list had eight rows of 13 syllables. He repeated a list until he could recite it twice without error.

Then he waited, from 20 minutes up to 31 days, and learned the same list again. His measure was savings, the share of the original learning time he did not need the second time. Savings of 100 percent would mean nothing was lost. Savings of zero would mean relearning took as long as learning. He ran 163 of these paired tests and reported them in Memory: A Contribution to Experimental Psychology.

What were Ebbinghaus's actual numbers?

The table shows his savings at each delay beside a modern replication. In 2015 Jaap Murre and Joeri Dros repeated the experiment in PLOS ONE with one subject, who learned 70 lists by the same method.

Time since learningSavings, EbbinghausSavings, Murre and Dros replication
20 minutes58.2%47.2%
1 hour44.2%37.3%
9 hours35.8%27.6%
1 day33.7%31.7%
2 days27.8%23.0%
6 days25.4%16.8%
31 days21.1%4.1%

What the numbers show

Look at the shape. Ebbinghaus lost more in the first hour than in the following month. The replication matched him closely except at 31 days, where the new subject's savings were much lower. An earlier German replication by Heller, Mack and Seitz, which Murre and Dros also report, found the same pattern in two more subjects.

The replication also shows a small rise at one day. Murre and Dros write that the curve 'most probably shows a jump upwards' at 24 hours. They connect it to research on sleep and add that their data cannot show sleep is the cause.

Is the forgetting curve exponential?

Diagrams often show a smooth exponential curve with a tidy formula. Ebbinghaus fitted his own data with a logarithmic formula and called it a shorthand statement of his results, with no claim to be a general law. In 1991 John Wixted and Ebbe Ebbesen argued that a power function describes forgetting better than an exponential, Ebbinghaus's data included. Murre and Dros found that a power function and a sum of exponentials fit almost equally well. The steep-then-flat shape is secure. The exact equation is still argued over.

Do you really forget 70 percent in 24 hours?

Training slides often claim that people forget about half of new information within an hour, 70 percent within a day and 90 percent within a week, and credit Ebbinghaus. The claim goes wrong in three ways.

  • He measured relearning time, and the claim is about facts recalled. Savings of 33.7 percent after a day means relearning took about two thirds of the original time. It does not mean two thirds of the content had gone.
  • His material was meaningless by design. He noted that stanzas of Byron's Don Juan took about a tenth of the repetitions he estimated for the same number of nonsense syllables. A day later, relearning the stanzas took about half the original work, where 12-syllable lists took two thirds.
  • The data came from one person learning by rote. The shape of the curve generalizes. The percentages do not.

How much will you forget?

How much you will forget of a book or a lecture depends on what it means to you, how well you learned it and what you do afterward. There is no universal percentage. The shape does hold outside the laboratory. In 1939 Herbert Spitzer had 3,605 sixth graders in Iowa read a 600-word article and tested groups after different delays. Scores fell as the delay grew and had leveled off by day 21, as a 2006 review by Roediger and Karpicke describes. Our guide on how to remember what you read covers the practical side.

What slows forgetting?

  • Meaning. Material that connects to what you already know is learned faster and held longer, as the Don Juan stanzas showed.
  • Retrieval. In a 2006 experiment by Henry Roediger and Jeffrey Karpicke, students who read a passage in four study periods lost 52 percent of what they could recall over the following week. Students who read it once and recalled it three times lost 14 percent. This is the testing effect.
  • Spacing. Reviews separated by gaps protect memory better than the same reviews packed together. In a 2008 study of 1,354 people by Nicholas Cepeda and colleagues, a well-timed review raised final recall by 64 percent over an immediate one.
  • Sleep, probably. Research on sleep and memory began in 1924, when Jenkins and Dallenbach reported better recall of nonsense syllables after sleep than after the same time awake. A 2017 review by Susan Sara still calls the role of sleep in memory 'controversial and elusive'.

How does review flatten the forgetting curve?

Ebbinghaus saw this too. He wrote that lists learned twice 'fade away much more slowly than those which have been learned but once'. When he relearned nine 12-syllable lists on six successive days, the work fell every day. The repetitions needed were 158, 109, 75, 56, 37 and 31.

Each review restores the memory, and the next decline is slower, so reviews can sit further and further apart. That is the logic of spaced repetition. The familiar diagram of several curves, each flatter than the last, illustrates this idea. It is not a plot of data Ebbinghaus reported.

The practical version is short. Recall the material from memory soon after learning it, again after a few days, and again after a few weeks. Recalling works better than rereading at each step, which is the case for active recall.

Where Generalist fits

Generalist: Micro Learning is built around this pattern. Its lessons take 3 to 7 minutes, and questions from lessons you completed come back later in short review sessions. You can browse mind and psychology or any of the other 15 subjects on the site.

Common questions

What is the Ebbinghaus forgetting curve?

It is the curve Hermann Ebbinghaus published in 1885 after testing his own memory for lists of nonsense syllables. It shows retention dropping quickly in the first hours after learning and then leveling off over days and weeks.

How much do we forget after 24 hours?

There is no fixed figure. Ebbinghaus needed about two thirds of the original time to relearn meaningless syllables after a day, a savings score of 33.7 percent. Meaningful material is retained better, and the popular claim of 70 percent lost in a day does not come from his data.

Who discovered the forgetting curve?

The German psychologist Hermann Ebbinghaus. He ran the experiments on himself in 1879 and 1880 and published them in 1885 in the book translated into English as Memory: A Contribution to Experimental Psychology.

How do you beat the forgetting curve?

Review before the memory has faded completely, and review by recalling. Test yourself on the material within a day or so, again after several days, and again after a few weeks. Each recall slows the next round of forgetting.

Is the forgetting curve still accepted?

Yes. A 2015 replication in PLOS ONE reproduced Ebbinghaus's curve closely with a new subject. Researchers still debate which mathematical function fits it best and how much the result generalizes from nonsense syllables to meaningful material.

Sources