Why you forget what you study (and why it isn't your fault)

You really studied. You highlighted, you summarized, you closed the notebook with a sense of duty done. Four days later someone brings up the subject, and the answer doesn't come. In its place comes the suspicion: weak memory, goldfish focus, maybe you're just not cut out for this.

It's none of that. Forgetting is the brain's default behavior, of every brain. Since 1885 there has been measurement showing that memory of freshly studied material plummets in the first hours and the first days, and then almost stops falling. The drop is steep, predictable, and democratic: it happens to the A student and to the C student alike.

The difference between those who remember and those who forget isn't talent: it's the calendar. Material is like people: whoever you never run into again becomes a stranger, no matter how good the first conversation was. Those who retain don't have a special memory; they have scheduled reunions. Below: the original 1885 measurement, the viral figure it never actually stated, and the minimum schedule to book that first reunion starting today.

The man who memorized nonsense syllables to measure his own forgetting

Hermann Ebbinghaus spent years of the 1880s doing something that sounds like punishment: memorizing lists of nonsense syllables (things like WUX, CAZ, BOK) and timing how much effort it took to relearn them later. Alone. He was the only subject in his own experiment.

The syllables were meaningless on purpose: a real word arrives with associations dangling from it, and he wanted to measure raw memory, unaided. And his measure wasn't "how much do I still remember," but savings: the economy of relearning. If learning a list cost ten minutes and relearning it the next day cost six and a half, he "saved" about a third of the original work.

The numbers from Über das Gedächtnis (1885): relearning after 20 minutes saved about 58% of the effort; after one hour, 44%; after one day, 34%; after one month, 21%. In his own words, one hour after learning, "forgetting had progressed so far that half the original work had to be spent again." Translated to the next day: about two-thirds of the initial work lost in 24 hours, on the worst material imaginable.

The pattern he described (forgetting "very rapid at the start and very slow at the end") became the famous forgetting curve. And it isn't old-textbook folklore: in 2015, Murre and Dros reran the experiment, 130 years later, with the same protocol. The curve came out with the same shape: 58.8% savings after 20 minutes, 35.6% after one day, 4.1% after one month. Even a tiny anomaly (a slight breather at the one-day point) showed up across both centuries.

This whole story, with the reran experiment and the math on what's left of two hours of studying after a week, is in the video:

"We forget 90% in 24 hours": the viral figure is wrong, and the real one already frightens

If you've ever looked up the forgetting curve, you've probably run into the line "we forget 90% of what we learn in 24 hours." It lives in training slides, in threads, in exam-prep posts. And it appears nowhere in Ebbinghaus's study.

The mistake is twofold. First, the real 1885 figure is different: by the savings measure, the loss in 24 hours was about two-thirds (a lot, but not 90%). Second, and more important: savings measures the time economy of relearning nonsense syllables, in a single person. You can't convert that into "the percentage of the lecture you still remember."

With material that makes sense, the drop is far gentler. In Roediger and Karpicke's (2006) experiment with prose passages, people who only read the text still remembered 54% of it two days later, and 42% after a week. That's the "read it and walked away" scenario, and even so it's a long way from the 10% the legend predicts.

The point that matters: you don't need to inflate the number to take the problem seriously. Losing half of what you studied within a few days, with honest effort invested, is already reason enough to change methods. Anyone who needs the "90%" to convince you is usually selling something.

Forgetting the material after studying isn't a lack of discipline: it's the method you were handed

School demands retention (the exam is three weeks away) and teaches absorption: attend the lecture, read the chapter, run the yellow highlighter across it. Nobody taught you when to review; at most they said "review more," as if it were a matter of willpower.

In 2013, Dunlosky and colleagues evaluated ten study techniques in a review for Psychological Science in the Public Interest. High utility: testing yourself and distributing study over time. Low utility: rereading, highlighting, summarizing. The two most-used techniques on the planet sit on the bottom shelf of the evidence.

Rereading fools you because it produces fluency: on the second pass the text goes down easy, and the brain confuses "I recognize this" with "I know this." Robert Bjork (1994) describes exactly this trap (short-term performance mistaken for learning) and calls the conditions that make studying hurt now to pay off later "desirable difficulties."

This illusion has been measured. In Roediger and Karpicke's second experiment, the group that reread the text four times was the most confident it would remember a week later. And it was the one that remembered the least: 40%, against 61% for those who read once and tested themselves three times. As a proportion of what each group knew at the start, the rereaders forgot 52%; those who tested themselves forgot 14%.

Put bluntly: if a method leaves you more confident and less capable (and rereading demonstrably does both), it isn't a weak method, it's a treacherous one. The highlighter deserved to be retired as soon as tomorrow.

What the forgetting curve respects: reunions at the right time

The way out has two pieces, and both fit the word "reunion."

Piece one: spacing. The meta-analysis by Cepeda and colleagues (2006), with 839 assessments across 317 experiments, compared the same total study time massed versus distributed: 36.7% correct on the final test for those who massed, 47.3% for those who spaced. When the test came 8 to 30 days later, the gap exploded: 32.8% against 62.2%. Nearly double, without studying a single minute more. The authors' practical conclusion: separating sessions by at least a day is "extremely useful" for long-term retention.

Piece two: the reunion doesn't call for rereading, but for testing. In Roediger and Karpicke's first experiment, five minutes after studying, rereading won (81% against 75% recall). One week later, the tables had turned: 56% for those who tested themselves, 42% for those who reread. Those who tested themselves held for a week what the rereaders held for only two days.

And when to book the first reunion? Ebbinghaus never prescribed a calendar. The ready-made schedules that circulate under his name are apocryphal. What actually exists is Cepeda et al. (2008): with 1,354 participants and tests up to a year later, the optimal interval before the first review grows along with the deadline: somewhere between 10% and 20% of the time until the exam. Exam in a month? First reunion in a few days. Want to remember for years? Space it out by weeks.

The full mechanism (why this interval exists, how it grows, and how to apply it without turning into a spreadsheet clerk) is the subject of our spaced repetition guide, the natural sequel to this post.

What to do today: five steps

Nothing here requires an app, a spreadsheet, or heroic willpower. It requires a calendar.

1. Close the material and produce. At the end of today's session, notebook closed: write or say everything you remember. Did it hurt? Good. The pain of trying to recall is the training, not the symptom.
2. Book the first reunion for tomorrow. At least a day away from the first session: that's the floor the Cepeda (2006) meta-analysis points to for lasting retention.
3. Calibrate by the deadline. First interval around 10% to 20% of the time until the exam (Cepeda, 2008): exam in 30 days, first review in three to six days, not three times on the same Sunday.
4. At the reunion, test before you check. Question first, material after. Reread only what you got wrong.
5. Stretch the intervals with each success. Remembered it well? The next reunion can wait longer. That progressive growth is the backbone of spaced repetition.

Frequently asked questions

Is forgetting quickly a sign of weak memory?

No. The forgetting curve is the default behavior of every brain. The original 1885 experiment and the 2015 replication found the same shape. What separates those who remember from those who forget is the review schedule, not talent.

How much do we forget in 24 hours?

It depends on the material. In the worst measured scenario (nonsense syllables), about two-thirds of the learning effort is lost in a day; with text that makes sense, you still remember 54% two days later. The "we forget 90% in 24 hours" going around isn't in any study.

When should I do the first review?

Between 10% and 20% of the time until the exam, according to the Cepeda et al. (2008) experiment. Exam in a month? First review in three to six days. Want to remember for years? Space it out by weeks.

Do rereading and highlighting really not work?

For long-term retention, they're the lowest-utility techniques in the Dunlosky et al. (2013) review. The same time spent testing yourself with the material closed pays off more, even when it feels harder (precisely because it feels harder).

Where the manual method breaks down

With one chapter, this system runs on paper. With a whole semester (sixty topics, each studied on a different day, each calling for a different interval), the reunion schedule turns into a second job. This is where most people give up: not on studying, on the bookkeeping.

And here's another point, equally blunt: a review spreadsheet is a week-one hobby. It doesn't die from laziness; it dies because it doesn't know what you've already forgotten. It treats the topic you've mastered and the one slipping away with the same bureaucratic care.

That's exactly the problem we tackled in Sift, the app we built: you study, and Sprint schedules the reunions. Each item comes back on the date when the review pays off most, in the form of a question, not a reread. The calendar stops being your task and becomes infrastructure.