Forgetting curve
Also known as: Ebbinghaus forgetting curve
Newly learned information drops away fast at first, then levels off.
What it means
The forgetting curve describes how retention of learned material declines over time, falling steeply soon after learning and then decelerating so that later losses come more slowly. Ebbinghaus discovered it by memorizing nonsense syllables and measuring savings in relearning at varying delays, establishing memory as something that could be studied quantitatively. The rate of forgetting depends on how deeply the material was processed, how meaningful it was, and how it was rehearsed, while the precise mathematical form of the curve remains debated. Crucially, its later course can be flattened by spaced retrieval, which is why distributed practice and active recall improve long-term retention so reliably. It remains a foundational result and the empirical backdrop for the spacing and testing effects.
The original demonstration
The result traces to Hermann Ebbinghaus, whose 1885 monograph reported experiments he ran on himself over several years. He learned lists of nonsense syllables - consonant-vowel-consonant strings such as WID or ZOF, chosen to be roughly free of prior meaning - and then, after a set delay, relearned each list to the same standard. His measure was savings: the proportion of learning trials or time spared on the second pass compared with the first. Because a list mastered again in half the original effort has clearly left some trace, savings gave a graded index of retention even where free recall would have scored zero. Testing at delays from about twenty minutes to a month, he found that savings dropped sharply in the first hours and then declined much more gently, so that a memory surviving a day was far more likely to survive a week. The demonstration mattered as much for method as for content: it showed that memory, long treated as too private for measurement, could be brought under quantitative control. Ebbinghaus was, however, his own and only subject, and the material was deliberately stripped of meaning - two features that shape how far the original curve generalizes.
What the evidence shows
Later work has repeatedly recovered a steep-then-shallow retention function, but the details have proved harder to pin down than the textbook figure suggests. In 2015 Jaap Murre and Joeri Dros repeated Ebbinghaus's procedure closely, with a single participant relearning syllable lists at the same delays he had used. The original shape reappeared, and Ebbinghaus's century-old numbers held up well, which is a genuine vindication of a one-person study. What the modern record does not settle is the mathematical form of the decline. Wixted and Ebbesen argued in 1991 that forgetting across many tasks is better described by a power function than by simple exponential decay. Rubin and Wenzel later fit more than a hundred retention data sets with a large set of candidate equations and found that several forms - logarithmic, power, and exponential in the square root of time - fit comparably well, with no single function clearly winning. Averell and Heathcote, using a longitudinal design and formal model comparison, reported that forgetting levels off at a stable, non-zero floor rather than continuing toward zero, consistent with a durable residue of well-learned material. The curve is real and reproducible; its exact equation remains an open question.
Why it happens
Several mechanisms are offered for the shape, and they are not mutually exclusive. One tradition attributes forgetting to decay: an unused trace weakens with the passage of time. A competing account emphasizes interference, in which later and earlier learning compete at retrieval, so that what looks like time-based loss is partly the accumulation of competing memories. A consolidation view adds that memories stabilize in the hours and days after encoding, which would explain why early losses are large and survivors comparatively robust. A quieter but important point concerns the averaged curve itself. When individuals or items forget at different rates, pooling them produces a smooth, decelerating aggregate even if no single memory follows that path; the group curve can therefore exaggerate how orderly individual forgetting is. This is why some researchers caution against reading the classic curve as a direct portrait of one mind losing one memory. Depth of processing, meaningfulness, and rehearsal all move the function, largely by determining how strongly material is encoded in the first place and how many retrieval routes lead back to it. None of these accounts has fully displaced the others, and the honest summary is that the curve describes forgetting well while its underlying causes remain partly contested.
Using it in practice
The practical payoff of the curve is that its slope is not fixed. Two interventions reliably raise later retention. The first is distributed practice: spreading study across separated sessions rather than massing it. A large quantitative synthesis of verbal-recall experiments found spaced practice consistently outperformed massed practice, with the best gap growing longer as the target retention interval lengthens. The second is retrieval practice, or testing: attempting to recall material, rather than simply rereading it, strengthens later memory. A meta-analysis of the testing effect put the average benefit at roughly a moderate effect size, and controlled studies show its signature clearly - repeated testing can look worse than repeated study on an immediate check yet produce markedly better recall a week later. Both effects can be read as ways of catching a memory just as it begins to fade and renewing it, which flattens subsequent decline. The applied caution is not to over-literalize any particular schedule. The curve does not license a universal rule that review must fall on fixed days; optimal timing depends on the material, the learner, and how long the knowledge must last. What the evidence supports is the direction of travel: spaced, effortful retrieval beats crammed, passive review.
Limits and caveats
Several qualifications keep the curve from being over-applied. Its founding data came from one highly practiced subject learning meaningless syllables, and meaningful, richly connected material forgets more slowly and less regularly than nonsense strings do. The classic smooth curve is usually a group or list average, and averaging can impose a tidy shape on messier individual data. The endpoint is also not zero: well-learned knowledge, such as a language studied for years, can settle into a long-lasting store that resists further loss for decades, so the curve should not be extrapolated to eventual total loss. The measure matters too; savings, recognition, cued recall, and free recall yield different apparent rates, and a memory scored as gone on one test may still show on another. Finally, the curve describes the fate of material after a single episode of learning under laboratory conditions; everyday knowledge is usually re-encountered, which resets and reshapes the function repeatedly. None of this undoes the core observation that retention falls fastest soon after learning. It does mean the forgetting curve is best treated as a robust qualitative pattern and a spur to spaced, tested practice, rather than as a precise law with a single equation or a fixed timetable.
Examples
Without review, much of a lecture's detail is lost within a day or two, with the remainder decaying far more slowly.
A language-learning application schedules each vocabulary item for review at lengthening intervals, timing the next prompt for roughly when recall is about to fail; across a term this keeps far more of the words retrievable than a single concentrated study session would.
A compliance team finds that staff pass a safety module immediately but score poorly on an unannounced quiz a month later; inserting two short spaced quizzes between the training and the audit noticeably slows that decline.
In a retention study, recall of a word list measured at twenty minutes, one day, and one week shows most of the total loss occurring within the first day, with the curve nearly level thereafter - the steep early segment made visible.
A student who crams the night before an exam and one who reviews the same material in four short weekly sessions may score alike on the test, yet months later the spaced learner retains substantially more.
First described in Hermann Ebbinghaus (1885).
Key references
- Ebbinghaus, H. (1913). Memory: A contribution to experimental psychology (H. A. Ruger & C. E. Bussenius, Trans.). Teachers College, Columbia University. (Original work published 1885) doi.org/10.1037/10011-000
- Murre, J. M. J., & Dros, J. (2015). Replication and analysis of Ebbinghaus' forgetting curve. PLOS ONE, 10(7), e0120644. doi.org/10.1371/journal.pone.0120644
- Wixted, J. T., & Ebbesen, E. B. (1991). On the form of forgetting. Psychological Science, 2(6), 409-415. doi.org/10.1111/j.1467-9280.1991.tb00175.x
- Rubin, D. C., & Wenzel, A. E. (1996). One hundred years of forgetting: A quantitative description of retention. Psychological Review, 103(4), 734-760. doi.org/10.1037/0033-295X.103.4.734
- Averell, L., & Heathcote, A. (2011). The form of the forgetting curve and the fate of memories. Journal of Mathematical Psychology, 55(1), 25-35. doi.org/10.1016/j.jmp.2010.08.009
- 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. doi.org/10.1037/0033-2909.132.3.354
- Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Psychological Bulletin, 140(6), 1432-1463. doi.org/10.1037/a0037559
- Roediger, H. L., III, & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249-255. doi.org/10.1111/j.1467-9280.2006.01693.x