You read a chapter on Monday. By Friday, most of it has gone. Nothing is wrong with you: that is what memories do, and a German psychologist put a curve on it more than 130 years ago.
In the 1880s Hermann Ebbinghaus wanted to study memory the way a chemist studies a reaction: with material that means nothing, so that no earlier knowledge could help. He invented nonsense syllables such as WID and ZOF, consonant-vowel-consonant strings, and learned long lists of them. There was one test subject: himself. He was 29, and the work went on for about seven months, often with up to three sessions a day.
His method was clever. He did not ask "can I recall it?", because after a while the answer is simply no. Instead he measured savings: how much quicker it was to relearn the same list. If a list took 20 repetitions the first time and 10 the second time, the saving was 50%. He did this after gaps of about 20 minutes, 1 hour, 9 hours, 1 day, 2 days, 6 days and 31 days, and published the results in 1885 in Über das Gedächtnis ("On Memory").
The shape he found is the famous forgetting curve: a steep fall at first, then a long, slow tail. Most of the loss happens early. Murre and Dros, in 2015, repeated the experiment with one volunteer, a 22-year-old, using 70 lists of 104 syllables each and the same range of delays. They found the same overall shape, and they also found that the 31-day measurement was far lower than the rest.
A tidy way to describe forgetting is an exponential: the chance you can still recall something after t days is R = e−t/S. The number S is the memory's stability, in days. A memory with S = 1 falls to about 37% after one day. A memory with S = 10 takes ten days to get there.
Real memories are messier than this, and the toy below is not Ebbinghaus's data. It is a model built to show one idea: each time you recall something, its stability grows, and it grows most when you had almost forgotten it. In this model a review at 100% recall adds nothing, and a review at 40% recall adds a lot. That is the heart of the spacing effect, and experiments since Ebbinghaus have kept finding that spread-out study beats cramming for long-term memory.
Tap the chart to schedule a review on that day. Each review snaps the memory back to 100% and makes the next fall slower. Then try the presets: both the cram and the spaced plan use the same four reviews.
Dotted line: 50% recall. Tap a dot to remove that review.
With no reviews, the memory is below 5% after about three days (with S = 1 day, that is 3 days exactly, since e−3 ≈ 5%). Four reviews crammed into the first day barely help, because each one happens while recall is still near 100%, so there is almost nothing to strengthen. The same four reviews spread over two weeks leave the memory far stronger at day 30, because each one lands when it counts.
The "just in time" button reviews each memory the moment it drops to 50%. With the default strength it keeps the memory highest at day 30 of the three presets, and the gaps stretch out on their own: short at first, then days, then weeks. That growing gap is exactly what spaced-repetition flashcard apps schedule for you.
Ebbinghaus measured savings, not recall, and savings stay positive long after you can no longer recite anything. A model with one number S cannot capture that, nor the differences between people, topics and sleep. Meaningful material is also forgotten far more slowly than nonsense syllables; his lists were deliberately the hardest case. Treat the chart as a way to feel the idea, and the idea as well supported: forgetting is steepest right after learning, and a well-timed second look beats a fifth look in a row.
So next time a fact matters, do not reread it five times tonight. Test yourself tomorrow, then in a few days, then in a couple of weeks.