The Forgetting Curve: How Memory Actually Works

The forgetting curve often appears in charts with neat percentages, such as “half forgotten in an hour.” Those numbers do not describe a universal rate of memory loss. This article examines what Ebbinghaus measured, how later research treats spacing, and why the assessment date matters.

What Ebbinghaus Actually Measured (It Is Not What You Think)

Ebbinghaus was probably the first scientist to study memory with experimental rigor, and he did it using himself as the only subject. For months he memorized lists of nonsense syllables, things like “DAX” and “BUP”, chosen precisely because they carried no meaning that could help memory along. Here is the first nuance the summaries leave out: he was not measuring “how much you still remember.” He was measuring savings: how much less time it took him to relearn a list to perfection, compared with the original learning time. When you read that “we retain 44% after one hour,” the original datum means that relearning the list one hour later required 44% less effort, which is quite different from being able to recall 44% of the content on an exam. The curve describes the decay of a memory trace, but the absolute percentages do not transfer directly to your biology notebook.

The Curve Has Survived 130 Years of Scrutiny

None of this diminishes the discovery, quite the opposite. In 2015, Jaap Murre and Joeri Dros redid the original experiment from scratch, with one subject memorizing lists of nonsense syllables over weeks, following the 1885 protocol down to the stopwatch. The replication, published in PLOS ONE, found a curve remarkably similar to Ebbinghaus’s: a steep drop in the first hours, progressive flattening afterwards, and even a curious small detail that Ebbinghaus himself recorded, a slight improvement around the 24-hour mark, possibly connected to the role of sleep in consolidation. Few results in psychology survive 130 years and a direct replication with that kind of robustness. The shape of the curve, fast forgetting at first and ever slower afterwards, is one of the most reliable findings we have about human memory.

What varies, and varies a lot, is the steepness. Meaningful material, connected to what you already know, decays far more slowly than random syllables. That is why a list of disconnected terms crammed the night before evaporates, while a concept you genuinely understood holds on. The curve is universal in shape, not in numbers.

Why rereading deceives: experimental evidence

If forgetting is so fast, why do so many people feel they are learning when they reread the chapter? The experiment that best answers this is Henry Roediger and Jeffrey Karpicke’s Test-Enhanced Learning (2006). Students read short passages and then either restudied the material or took a recall test on it. Assessed five minutes later, the restudy group did better. Assessed one week later, the picture flipped completely: those who had been tested remembered substantially more.

A useful detail from that study is that the participants who restudied predicted they would do better. Rereading generates fluency, the text flows easily, and the brain interprets that ease as mastery. It is a metacognitive illusion measured in the lab: the strategy that feels best in the short term is the one that produces the least durable memory. The practical implication for a study tool is to support testing rather than only rereading: a practice quiz is not just a way to measure what you know, it is itself the event that strengthens the memory.

When to Review: What the Spacing Evidence Actually Says

Spaced repetition means returning to content after intervals. A fixed rule such as “1, 3, 7, and 21 days” oversimplifies the evidence. Nicholas Cepeda and colleagues’ meta-analysis, Distributed practice in verbal recall tasks (2006), pooled experiments favoring separate sessions over concentrating practice at once in the materials and conditions studied.

In another study, Cepeda and colleagues (2008) tested review intervals and assessment delays. The most favorable interval varied with the time until the test. This supports planning a return to the material around the test date rather than applying the same calendar to every goal. The exact experimental intervals are not an individual prescription.

And the most counterintuitive finding: overshooting the interval costs little; undershooting costs a lot. Performance declines gently when spacing runs past the optimal point, but it drops sharply when reviews are bunched up too early. In practice this inverts the instinct of the anxious student who re-reviews everything the next day: if your exam is far away, reviewing too early and too often is the most expensive way to feel like you are studying. That struggle to remember after days away from the material is not a sign of failure, it is exactly the condition under which a review strengthens memory the most.

Where the Curve Does Not Apply

To be honest with the evidence, it is worth saying what the forgetting curve does not describe well:

• The percentages are not laws. As we saw, Ebbinghaus’s numbers come from nonsense syllables and the savings method. Quoting “you forget 70% within 24 hours” as a universal statistic, as many articles do, is a misquotation.
• Procedural skills decay differently. Riding a bike, playing an instrument, solving a well-practiced class of problems: procedural memories are far more resistant than freshly learned declarative facts. The steep curve applies above all to new factual content.
• Understanding changes the slope. Content integrated into a structure you understand decays slowly. Applying spaced repetition to material you memorized without understanding is optimizing the wrong thing: comprehension first, retention second.

Applying spacing with SimulAI and avoiding mistakes

One possible workflow is to answer questions soon after studying, log mistakes, and schedule further rounds before the assessment. Use flashcards for definitions and tests for relationships and applications, checking the answer key against the source. Adjust intervals to what you can still retrieve; the platform does not calculate an individual forgetting curve.

Two uses worth avoiding are:

• Retaking the same quiz on the same day until scoring 100%. The feeling of progress is wonderful, but it is the short-term fluency from Roediger and Karpicke’s experiment: the real gain comes from retaking it days later, when remembering hurts. Generating new variations on the same source may help test the concept instead of a memorized answer, provided the new questions are checked against the source.
• Treating flashcards as reading. Flipping the card before attempting an answer turns active recall into disguised rereading. The effort of trying, even when you get it wrong, is the active ingredient; without it, the deck is just a chopped-up summary.

An Honest Schedule to Start With

• On study day: finish with a recall test, not a rereading. Close the material and write or answer from memory.
• Set the horizon: look at the exam date and distribute two to four reviews between now and then, at increasing intervals, with the first major gap somewhere around 10% to 30% of the total time until the test.
• Every review is a test: questions, flashcards answered before flipping, explaining out loud. Rereading is plan B, never plan A.
• Calibrate by error: retrieval that feels too easy means the interval was too short; a pile of errors means the gap ran too long or the foundation lacked understanding.

Forgetting some material is expected. Planning new retrieval opportunities can make review more useful than rereading on impulse. No chart predicts how much a particular person will forget after a fixed period; use actual answers to decide the next step.

References

• Murre, J. M. J., & Dros, J. (2015). Replication and Analysis of Ebbinghaus’ Forgetting Curve. PLOS ONE, 10(7), e0120644. journals.plos.org
• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380. psycnet.apa.org
• Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095–1102. journals.sagepub.com
• Roediger, H. L., & Karpicke, J. D. (2006). Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science, 17(3), 249–255. journals.sagepub.com