In the winter of 2011, a 22-year-old student in Amsterdam sat at a desk and read nonsense syllables aloud, thirteen to a row. He repeated each row until he could recite it without a mistake. Then he waited twenty minutes, a day or a month, and learned the same rows again. By February he had spent about 70 hours on it. He was repeating an experiment Hermann Ebbinghaus first ran on himself in 1879.
He got the same curve. The forgetting curve shows that we forget fastest right after learning, and then more and more slowly. In Ebbinghaus's own results, an hour after learning he had to spend half the original effort to learn the syllables again. After a day, about a third of the work was still saved. After a month, about a fifth. For anyone learning words, this has one clear consequence: the first review should come soon.
What Ebbinghaus actually measured
Ebbinghaus wanted to study memory without the help of meaning, so he used short syllables that mean nothing. In 1879 and 1880 he ran 163 tests on himself. Each time, he learned a set of rows until he could recite them twice without an error. Then he waited a fixed time (about twenty minutes, one hour, nine hours, one, two, six or 31 days) and learned the same rows again.
He did not count how many syllables he could still recall. He measured how much time he saved the second time. He called this "savings".
That detail matters, because the curve is often misquoted. It does not say that you keep a third of a word list after one day. It says that relearning the list took about a third less work than learning it did. Something you can no longer recall on demand can still be stored, and it comes back faster than it went in. In Ebbinghaus's words, forgetting is "very rapid at the beginning of the process and very slow at the end" (Memory, chapter 7).
The 2015 replication
Results from one man in 1880 deserve a second test. In 2015, Murre and Dros of the University of Amsterdam published one in PLOS ONE. Dros, the second author, was the only subject, just as Ebbinghaus had been. He followed the original method as closely as the old description allowed, with the same seven intervals and ten lists for most of them.
His curve looked very much like the original, and like an earlier German replication with two subjects. There was one clear difference. At 31 days, Dros had saved much less than Ebbinghaus. The authors could only guess at the reason. One possibility they mention: all his 31-day lists were learned early, in a short period.
The more interesting finding was a small bump. Across the old and new data, the curve is probably not perfectly smooth. Memory looks slightly better than expected at the one-day point, after a night has passed. The authors note that sleep research would predict exactly this, but they are careful: for this kind of experiment, sleep as the cause "remains to be established". In Dros's own data, there was no clear sign of the bump.
Sleep and the first review
Other research supports the idea that sleep helps. In a 2016 study in Psychological Science, 40 people practised foreign vocabulary until they knew it perfectly, then relearned it 12 hours later. Half learned in the morning and relearned in the evening of the same day. The other half learned in the evening, slept, and relearned the next morning.
The sleep group needed half as much practice to relearn the words. They also remembered them much better one week later and six months later.
Put these findings together and the curve stops looking like bad news. Forgetting is steepest in the first hours. A night of sleep seems to slow it down. And whatever you have lost comes back with less effort than it took the first time.
What the curve can't tell you
Three limits are worth knowing. Ebbinghaus and Dros each tested one person. Both used syllables with no meaning, which is not how you meet words in a real language: a word in a sentence has more to hold on to. And savings is a sensitive measure that picks up memories you could not recall when asked. So don't read exact percentages off the curve for your Spanish vocabulary. Read the shape.
What this means for you
- Review new words within a day. The first gap should be short, because that is where the curve falls fastest.
- Let a night fall between learning and the first review. Study new words in the evening and go over them the next morning.
- Make each gap longer than the last. Our guide to spaced repetition explains the schedule and why recalling beats rereading.
- Don't count a forgotten word as wasted time. Relearning it is faster than learning it was. That is exactly what Ebbinghaus measured.
- Give words meaning. Learn them inside short sentences rather than as single items on a list, so they don't behave like nonsense syllables. Our comparison of vocabulary learning methods shows how.
Qlango builds this shape into its lessons. A finished lesson comes back one day after you finish it, then after two days, four days, eight days and eight days again. The first review lands in the steep part of the curve, and the later ones move further apart. A lesson turns red when it is due, and examples you needed a hint for return later in harder questions.
Dros spent 70 hours on syllables that meant nothing, and his memory still behaved much like Ebbinghaus's had more than 130 years earlier. Your words mean something, which already puts you ahead. Learn them tonight, look at them again tomorrow morning, and let the curve work for you. If you want to know how many of those mornings a whole level takes, here is how long it takes to learn a language.
Frequently asked questions
What is the Ebbinghaus forgetting curve?
It is a graph of how memory fades over time, first measured by Hermann Ebbinghaus in 1879 and 1880 with nonsense syllables. It falls steeply in the first hour and the first day after learning and then flattens out, so most forgetting happens early.
How much do we forget after 24 hours?
In Ebbinghaus's experiment, relearning after 24 hours took about a third less work than the first learning. This is often misquoted as "we remember only a third". His numbers came from meaningless syllables and one person, so they can't be applied directly to real words. The shape of the curve is the reliable part.
How do you beat the forgetting curve?
Review soon after learning, then at growing intervals, and try to recall each word before you look at the answer. Sleep between learning and the first review also helps: in a 2016 study, it halved the practice people needed to relearn foreign words.
Has the forgetting curve been replicated?
Yes. A 2015 replication at the University of Amsterdam, published in PLOS ONE, produced a curve very close to the original. It also found signs of a small upward jump at the one-day point, which may be linked to sleep, though the authors say this still needs to be shown.
