The Forgetting Curve: Why You Forget What You Study

The forgetting curve is a graph showing how quickly information fades from memory when you do not review it. It drops steeply in the hours just after you learn something, then levels off. Hermann Ebbinghaus mapped it in the 1880s, and it is the single best explanation for why a topic you understood perfectly on Monday feels blank by Thursday.

This post is about the why, not the what to do. Understanding the shape of the curve is what turns spaced review from a chore someone recommended into something you can see the logic of. Here is what the curve is, who discovered it and how, what its shape actually means, why memory decays that way, and what happens to the curve when you review.

What is the forgetting curve?

The forgetting curve is a line describing how much of a newly learned set of material you can still retrieve as time passes without any review. Time runs along the bottom, from the moment you finish learning; retention runs up the side. The line starts high and falls, but the important detail is that it does not fall evenly. It plunges at the start and then bends, settling into a long, shallow tail.

That bend is why the curve is worth knowing about at all. If forgetting were a steady trickle, it would not matter much when you scheduled a review. Because most of the loss is packed into the beginning, timing becomes the lever. The curve reframes forgetting as a predictable physical process with a shape, rather than a random personal weakness, and once something has a shape you can plan around it.

Who discovered the forgetting curve?

The German psychologist Hermann Ebbinghaus, who pioneered the experimental study of memory, began his memory studies at the University of Berlin around 1879 and published them in 1885 as Über das Gedächtnis, which reached English readers as Memory: A Contribution to Experimental Psychology. Before him, memory was treated as something to philosophise about rather than measure. He was, as the record of his work notes, his own and only experimental subject.

His method was deliberately austere. He invented nonsense syllables, consonant and vowel and consonant strings such as WID or ZOF that carried no prior meaning, and built a stock of around 2,300 of them. He would draw a set at random, read them aloud to the beat of a metronome until he had them, and then wait. The point of stripping out meaning was to remove the advantage that familiar words get from everything you already know, so that what he measured was closer to raw retention.

He also chose a clever way to measure what was left. Rather than asking how many syllables he could still recite, which collapses to zero quickly and tells you nothing after that, he measured savings: how much less effort it took to relearn a list than it had taken to learn it the first time. A list that felt completely gone could still be relearned faster than a fresh one, which showed that a trace survived below the level of conscious recall. Savings let him keep measuring long after plain recall had bottomed out.

What does the shape of the curve actually mean?

Read it in three phases rather than as one number. Be sceptical of the specific percentages that circulate online alongside this curve, because most are repeated without a source and the real figures depend heavily on the person, the material, and how well it was learned to begin with. The shape is the robust finding; the exact numbers are not.

The practical reading is counter-intuitive. Most students save revision for the week before an exam, which is precisely the flat part of the curve where each hour buys the least, and they skip the day after the lecture, which is the steep part where the same hour buys the most. The curve says the cheapest memory you will ever buy is the one you top up early.

It is fair to ask whether a solo experiment from the 1880s still stands. It largely does. In 2015, Jaap Murre and Joeri Dros published Replication and Analysis of Ebbinghaus' Forgetting Curve in PLOS ONE. Using the same savings method, one subject spent about 70 hours learning lists of nonsense syllables and relearning them after delays ranging from 20 minutes to 31 days. The results came out similar to the original data, and the authors concluded that the forgetting curve had indeed been replicated. They also found it is not completely smooth, and most probably shows a jump upwards starting at the 24 hour data point.

Why do we forget what we learn?

Partly because a fresh memory is physically fragile. When you first learn something the trace is unstable, and it takes time and repetition before it settles into a durable form. Sleep does a lot of that consolidation work, which is one reason a night of proper rest after studying is not wasted study time.

But a large share of what feels like forgetting is really retrieval failure. The information is still in there, as the savings method demonstrated, and what you have lost is the path back to it. This is why a hint or a first letter can suddenly restore an answer you would have sworn was gone, and it is why practising the act of retrieval, rather than the act of reading, is what strengthens the route.

There is also a trap in how confident you feel. Re-reading notes makes the page feel familiar, and your brain reports that familiarity as knowledge, so you close the book believing you have learned it. That is the fluency illusion, and it is the reason highlighting feels productive while doing little for retention. The curve keeps falling underneath the good feeling. Finally, Ebbinghaus deliberately used meaningless syllables, which forget fastest of all. Material you genuinely understand and can connect to things you already know sits on a gentler curve than material you memorised as an arbitrary string, so building meaning is itself a defence against decay.

How do you beat the forgetting curve?

You do not stop the curve, you reset it. Each time you successfully retrieve something, the memory is strengthened and the next decline starts from a higher point and falls more gently. Repeat that a few times and the curve flattens until the material barely decays at all between checks. This is the entire mechanism behind spaced repetition: reviews are placed to interrupt the drop while it is still steep, and the gaps widen because each successful pass earns you a longer safe interval.

Two conditions matter. The review has to be retrieval, meaning you produce the answer from memory before checking it, since re-reading does not reset the curve in the same way. Our comparison of active recall versus spaced repetition covers how those two ideas fit together, and the wider set of methods that improve retention puts them alongside the others worth using.

The second condition is timing, which is where this post hands off. Deciding the actual days on which those reviews land is a separate job, and our guide to building a spaced repetition schedule lays out a concrete plan you can follow. If you prefer to let a stack of cards do the scheduling instead of a calendar, the Leitner system is the classic low-tech way to get widening intervals without tracking dates at all.

How GeniusPal helps

The curve is only useful if you have something to test yourself with, and building that material is the part most people never get around to. GeniusPal handles that step: upload your notes, a PDF, or a chapter, and it turns the content into a quiz, a flashcard deck, a mind map, or a summary. Because the output is question-shaped rather than a wall of text to re-read, it pushes you toward retrieval, which is the kind of review that actually resets the curve.

Be clear on the boundary, though. GeniusPal does not run a spaced repetition scheduler and it will not remind you when a topic is due. It makes the practice material; the timing is yours to manage, whether through your own routine, a calendar, or a dedicated spaced-repetition app if you want the scheduling automated. The free plan covers 2 study sets for the lifetime of the account, which is enough to see whether generated questions are a format you will actually revisit. Ebbinghaus needed 70 hours to draw his curve, but you only need to act on the finding: test yourself early, test yourself again before you forget, and let the gaps grow.