How fast we forget

In one line: we forget fast at first and then more slowly, and spreading practice out takes less effort than doing it all at once.

What he did

Hermann Ebbinghaus was a German psychologist. In the early 1880s, working alone, he set out to measure forgetting. He was his own and only subject.

He made up lists of nonsense syllables, such as DAX, BUP and ZOL. He chose nonsense on purpose, so that nothing he already knew could help him. He learned each list until he could recite it perfectly. Then, after a gap of anything from 20 minutes to a month, he learned it again.

He didn’t measure how much he could recall. He measured how much time he saved the second time round. If a list took 10 minutes to learn and 4 minutes to relearn, he had saved 60% of the effort. This “savings” method could pick up memory even when he couldn’t recall a single syllable.

He did this over and over, with thousands of syllables, for years, and published the results in 1885.

What he found

• Forgetting is fastest at the start. After 20 minutes, he saved about 58% of the effort. After an hour, about 44%. After a day, about 34%.
• Then it slows right down. After 6 days he still saved about 25%, and after a month about 21%. Most of the loss happened in the first day.
• Spreading practice out needed far fewer repetitions. For one list, 38 repetitions spread over three days did as much as 68 repetitions in a single sitting.
• Memory can hide. Lists he couldn’t recall still took less time to relearn. Not being able to remember something isn’t the same as having lost it completely.

How much should you trust it?

A famous, careful classic: trust the shape of the curve, not the numbers.

• It was one person, testing himself. Nobody else’s memory was measured.
• Nonsense syllables are the opposite of most healthcare learning. Material that means something to you fades much more slowly.
• The numbers are savings in relearning time, not the share you can remember. Claims such as “you forget 70% within a day” often misuse them.
• The shape has held up well. A modern team repeated the experiment in 2015 and got a very similar curve.

If you need to convince someone

In our words (a paraphrase, not a quotation): most forgetting happens soon after learning. If we never come back to what we teach, we lose a lot of it, however good the session was.

What it means for you

1. Plan for forgetting. Forgetting is fastest soon after learning. Clinical material fades more slowly than nonsense syllables, but a single session with no follow-up still loses a lot. That’s a property of memory, not a sign of poor teaching.
2. Follow up soon, then again later. A quick question in the days after a session helps more of it last. Add later ones for things people need all year.
3. Ask before you re-teach. If people can’t answer a question about something from last month, they probably haven’t lost it all. A prompt or a single case may bring it back faster than teaching it again from scratch.

How it appears in Teaching That Lands

Session 2 includes a nod to Ebbinghaus: a sheet of made-up words with silly hospital meanings, which half the room re-reads and half test themselves on. After a surprise retest, the facilitator draws the forgetting curve: most of it goes fast, then levels off, and each time you pull it back out, it fades more slowly. That’s one reason the sessions are a week apart, and why the made-up words are tested again at the start of Session 3.

In Session 3, “forgetting is the default” comes back when participants plan what their own learners will revisit at 30, 60 and 90 days. More on the spaced practice page.

The small print

• Ebbinghaus summed up his results with a formula based on the logarithm of time, not, as often stated, an exponential decay. Both describe a curve that falls steeply and then flattens.
• The English translation (Ruger and Bussenius, 1913) is free to read online and is in the public domain.
• Murre and Dros (2015, PLOS ONE) repeated the experiment with one subject, and found a very similar curve. This supports using the shape, not the exact numbers.
• Studies of medical knowledge show much slower forgetting than Ebbinghaus found. One review (Custers, 2010) estimated that roughly two-thirds to three-quarters of basic science knowledge was still there after a year, falling to about half after two years.
• Spaced practice and the forgetting curve are linked: see the review of Cepeda et al. (2006).