# Why Students Forget After Exams

> Exam knowledge fades within months because the review schedule ends on exam day. Here is what the research says survives, and how to keep more of it.

Source: https://getmemset.app/blog/why-students-forget-after-exams  
Published: 2026-09-29

---

Every semester ends the same way. A week of intense revision, one morning of concentrated recall, and then a strange silence. Nobody asks you about the Krebs cycle or the Treaty of Westphalia again. The notes go into a drawer, the textbook goes back on the shelf or back to the seller, and the subject quietly falls out of your life. Six months later someone mentions it at dinner, and you realise you can name the course but not much inside it.

Students call this the brain dump, and they usually say it with a shrug. It feels like a law of nature. You learn, you pass, you forget, and the cycle starts again in September.

There is nothing natural about it. The cycle is a scheduling outcome, and a fairly predictable one. A course is a sequence of dates that ends on exam day. Every review you do is pointed at that date. When the date passes, the reviews stop, and memory does what memory does when reviews stop. For years I assumed my problem was how much I studied. It took a long time, and a spaced-repetition habit I eventually turned into [a review planner](https://getmemset.app), to see that the problem was where the schedule ended.

This article is about why the post-exam collapse happens, what the research says actually survives, and what it takes to keep a subject alive after the last date on the calendar.

![A 3D-rendered blank desk calendar with one cell ringed in blue, and a glossy translucent blue ribbon arcing over it that peaks above the ring, then descends and dissolves into drifting cubes and dust](https://getmemset.app/blog/why-students-forget-after-exams.webp)

## The exam is the last date on the schedule

Look at how any course is built. Lectures run for twelve or fifteen weeks. Assignments land every fortnight. Revision compresses into the final two weeks. Then the exam. Every piece of that structure points at one morning, and the structure ends when the morning does.

Your memory has no idea the course is over. It registers two things only: whether something was retrieved recently, and how much effort the retrieval took. During revision week, retrieval is constant. On exam morning it peaks. The next day it drops to zero, and stays there.

So the brain dump is a misleading name. Nothing gets dumped. What happens is an absence of retrieval, repeated every day for months. **The reviews stopped**, and that is the whole mechanism.

I find this framing useful because it moves the blame. The student did nothing wrong on exam day. They followed a schedule to its end. The schedule was designed to produce one strong performance on one date. It succeeded, and then it had nothing else to say.

## What the curve does after the peak

The shape of what happens next has been measured for over a century. Hermann Ebbinghaus first plotted [the forgetting curve](https://getmemset.app/learn/the-forgetting-curve-explained) in the 1880s by memorising nonsense syllables and testing himself at intervals. The drop is steep at first and then flattens. In 2015, Jaap Murre and Joeri Dros [replicated the original experiment](https://doi.org/10.1371/journal.pone.0120644) with modern controls and got almost the same curve.

The detail that matters for exams is where the curve starts. Exam day is the peak. Everything you knew that morning sits at its highest point, and the curve begins its steepest descent from exactly there. A week of concentrated revision builds a tall peak, and a tall peak has a long way to fall.

Compare that with material you learned early in the semester and kept meeting in later weeks. Its curve got interrupted over and over. Every time you needed a concept from week two to understand week six, you retrieved it, and each retrieval reset the curve at a slightly flatter slope. That material tends to survive the summer. The material from the last two chapters, learned once and tested once, usually does not.

Same student, same course, same exam. The difference is how many times each piece was pulled back out of memory before the schedule ended.

## What survives fifty years

The most striking evidence about post-exam forgetting comes from studies that waited a very long time before asking.

In 1984, Harry Bahrick tested [nearly 800 people on the Spanish they had learned in school](https://doi.org/10.1037/0096-3445.113.1.1), anywhere from one to fifty years earlier. Most of the loss happened in the first few years. After that, whatever remained stayed remarkably stable for decades. Bahrick called this stable residue **permastore**. How much reached it depended on how much had been learned in the first place, and on how that learning had been spread out. Spanish spread across more courses and more years was more likely to make it.

A few years later, Martin Conway and colleagues tested [former psychology students anywhere from a few months to twelve years](https://doi.org/10.1037/0096-3445.120.4.395) after they had finished a cognitive psychology course. Names and specific findings faded steadily. Research methods and broad concepts held up much better. The kind of knowledge mattered far more than the years that had passed: details decayed, while methods and concepts stayed.

Both studies point the same way from different angles. What lasted was learning that had been spread out over years, and the conceptual core rather than the details. A single revision sprint aimed at one date produces neither.

> Memory does not know the course is over. It only knows the reviews stopped.

## The one-third rule

Medicine is the field that has measured this most carefully, because a medical student forgetting anatomy has consequences. Eugène Custers [reviewed the retention studies](https://doi.org/10.1007/s10459-008-9101-y) on basic science knowledge in medical education and found a consistent figure. Students lose roughly a quarter to a third of what they learned within a year of the exam, and about half within two years.

Two things in that review are worth sitting with. First, retention is roughly proportional to what was learned. Students who learned more also kept more in absolute terms, so the effort was not wasted. Second, the loss is front-loaded. The first year after the exam does most of the damage, which is exactly the window in which nobody is reviewing.

If that is what happens in a field with strong incentives to remember, the numbers for an elective you took to fill a timetable are probably worse. I have no study for those. I do have my own experience, and the residue from courses I never touched again is close to nothing.

## The gap should match the horizon

There is a second reason exam-shaped studying fails after the exam, and it is about the size of the gaps between reviews.

In a large study published in 2008, Nicholas Cepeda and colleagues [varied the spacing between study sessions](https://doi.org/10.1111/j.1467-9280.2008.02209.x) and then tested people after delays ranging from a week to nearly a year. The best gap was never fixed. It grew with the delay. If the test was a week away, a one-day gap worked well. If the test was a year away, the best gap was about three weeks. As a rough rule, the ideal spacing came out somewhere between five and twenty percent of the time you wanted the memory to last, with the fraction shrinking as the horizon grew.

Now look at a revision schedule. The gaps are a day, sometimes hours. That spacing is tuned for a test about a week away, which is precisely what the exam is. The schedule works because it fits the horizon. The trouble is that the horizon is wrong. A student who wants to still know the subject in two years needs gaps measured in weeks and months, and nothing in the course ever asks for those.

So the post-exam collapse is baked in twice. The reviews stop on exam day, and the reviews that did happen were spaced for a horizon that ended on exam day. **The gap follows the horizon**, and the course only ever had one horizon in mind.

## Why we do it anyway

None of this is a character flaw. The incentives point one way and the science points the other, and incentives usually win.

A course rewards performance on a date. Nobody measures whether you know the material next year, so the course has no reason to care. Revision week feels productive because performance climbs in front of your eyes. Spaced review over the following months feels pointless because there is no test to show the gain, and the gain is invisible by design. The whole benefit of spacing is that it stops something from happening, and you cannot watch something not happen.

There is also a simpler reason. After the exam, you are tired of the subject. Two weeks of intense contact with any material produces a kind of allergy, and the last thing anyone wants in July is to open the statistics notes again. The dump is partly a relief.

I am not going to argue that everyone should keep every course alive forever. That would be its own kind of failure. Some subjects really are done the day the exam ends, and letting them go is fine. The problem is the subjects that are not done: the anatomy you will need in the clinic, the algorithms you will need in the interview, the language you will need on the trip, the statistics you will need the moment you start your own research. Those get dumped along with everything else, because the schedule made no distinction.

## Keeping a subject alive without a second job

If you want a subject to outlast the exam, the fix follows directly from the research, and it is smaller than it sounds.

First, decide what actually needs to survive. This is most of the work and it takes an afternoon. Out of a whole course, the part you will need in two years is usually a short list: the core concepts, the methods, the few facts everything else hangs on. Conway's twelve-year study is a good guide here. Concepts and methods have a real chance at permanence. Names, dates, and the specific figures from lecture nine mostly do not, so unless you need them, let them go.

Second, put that short list on a schedule that continues past the exam and whose gaps grow. Daily gaps are the wrong shape for this. Something closer to a week, then two, then a month, then three months. This is the ordinary shape of [spaced repetition](https://getmemset.app/learn/spaced-repetition), and the only unusual part is starting it the week after the exam rather than the week before.

Third, keep the material where it already is. The most common way this plan dies is that someone decides to rebuild the whole course as flashcards first, and never finishes. The notes you already have are fine. A review is a pointer back to a page, a chapter, a problem set, plus one honest attempt to reproduce it from memory before you look.

Fourth, cap the daily load. A schedule that spans several subjects will eventually stack reviews on the same day, and the first time you open it and see forty items due is the last time you open it. A hard ceiling per subject per day, with the overflow rolling forward, keeps the habit survivable through the weeks when life gets in the way.

That is the whole method. A short list, growing gaps, pointers instead of rebuilds, and a daily limit. It takes perhaps ten minutes a day, and it is the difference between a course that leaves a residue and a course that leaves a transcript entry.

## The residue is the point

Bahrick's subjects kept what they kept for fifty years. That is a long time to hold onto a school subject, and it happened without anyone trying, simply because the learning had been spread out enough to reach the stable part of the curve. Everything else drained away over the first few years.

The exam itself is fine. It is a useful deadline and a fair measurement of a moment. The mistake is treating it as the end of the schedule for material that has to last, and then being surprised in October when the material is gone.

If you want a way to keep the reviews going after the last date on the calendar, with growing gaps and a daily limit that keeps revision from becoming a second job, that is what [Memset](https://getmemset.app) is built for.