What Ebbinghaus’s forgetting curve really shows

Hermann Ebbinghaus was a German psychologist. In 1885, in his book Über das Gedächtnis, he used himself as the only subject. He learned lists of meaningless syllables, so that nothing in the list could help him remember it. Then he waited for a set delay and relearned the same list.

What he wrote down was the time saved on the second learning compared with the first. He repeated the experiment with seven delays, from 20 minutes to 31 days (Ebbinghaus, 1885). Join the points and the curve drops fast at the start, then flattens out. That is the famous forgetting curve. It is not a magic law. It is a series of measurements on one person, with very odd material, but it gave a shape that almost every psychology course still shows.

Ebbinghaus did not count the syllables he remembered. He measured a saving, meaning the time gained when he relearned the list. Imagine that the first time he took a certain time, and the next day he took half as long. The saving is then one half. That does not mean he remembered half of the list.

He could be unable to recite the list and still save time, because he relearned familiar syllables faster. A saving of 33.7% after one day (Ebbinghaus, 1885) therefore does not mean “you keep 33.7% of your course”. It shows that a trace is left, even when you can no longer recite anything. This nuance explains why the numbers on the curve so often travel with wrong captions. When you read a percentage under a graph, always ask yourself what it really measures.

More than a century later, Murre and Dros ran the experiment again. A single subject, Dros, learned lists for 70 hours in total, then relearned them after 20 minutes, 1 hour, 9 hours, 1 day, 2 days, 6 days or 31 days (Murre and Dros, 2015). The result looks like Ebbinghaus’s. Savings drop fast, then slow down, as the table below shows.

The interesting point is the rise at 24 hours. Dros’s saving goes from 27.6% at 9 hours to 31.7% at one day. The authors see this as probably an effect of sleep (Murre and Dros, 2015). Another gap is that after 31 days Dros saves only 4.1%, against 21.1% for Ebbinghaus. Two people, two ways of testing, and already the values differ. That is why the curve gives a general shape, not a scale to apply to your own brain.

Keep the shape, not the percentages. Forgetting is fast at the start, then it slows down. That is why going back over an idea soon after learning it costs almost nothing, while going back later takes an effort that memory seems to reward.

Another study shows what effort changes. In Roediger and Karpicke’s experiment, between the five-minute test and the one-week test, those who had reread four times forgot 52% of what they knew, those who had quizzed themselves 14% (Roediger and Karpicke, 2006). And for vocabulary, retesting a word already known had a large positive effect on delayed recall, while studying it again changed nothing (Karpicke and Roediger, 2008). For finals, AP exams or the SAT, the lesson is simple. Instead of rereading once more, make yourself say the idea just before it fades.