Spaced repetition is the plan of coming back to the same material after gaps, rather than going over it several times in one sitting. The spacing effect behind it is among the oldest findings in psychology; the practical question is how to schedule revision, meaning how long to wait between reviews. The studies say it depends on how far away the test is, and that waiting a little too long does much less harm than coming back too soon.
Why spacing and self-testing beat rereading is covered in our guide to how to study effectively. This one covers the forgetting curve, the gaps each experiment found, the Leitner system and flashcard apps, and a revision calendar that is ours.
EdCitation publishes this guide. Every study was read in the paper itself in September 2026, except Kang (2016) and Kornell (2009), of which we read only the abstracts; of Ebbinghaus's book we read the two chapters cited. The references were built from their DOIs in Cite a source, the whole list went through Verify references, and Find sources is where to search for more.
What is spaced repetition, and why does it work?
Spaced repetition is reviewing the same material on separate occasions with time between them; the benefit it produces is called the spacing effect, or distributed practice. Its opposite is massed practice: the same number of repetitions packed together, as in cramming.
Ebbinghaus (1913), in the English translation of a book first published in 1885, reported that 38 repetitions spread over the three days before a test did as well as 68 made on the day before, and judged a suitable spread of repetitions over time "decidedly more advantageous" than massing them. The evidence was one person learning nonsense syllables.
Kang's (2016) abstract says hundreds of studies show spaced encounters beat massed ones for long-term learning, that adding tests increases the gain, and that the benefit reaches problem solving and new situations. Why spacing works is not settled; Cepeda et al. (2009) write that no consensus has emerged.
What does the forgetting curve actually show?
The forgetting curve shows that most of what is lost goes quickly, within hours, and the rest goes slowly for weeks. That shape is the reason review has to be scheduled at all.
Ebbinghaus's curve
Ebbinghaus measured forgetting by savings: how much less time relearning a list took than first learning it. An hour after learning, half the original work had to be done again; after 24 hours about a third was still saved, after six days about a quarter, after a month about a fifth. The subject was Ebbinghaus alone, learning lists of 13 nonsense syllables.
Savings is not the share of a list a person can recall. Murre and Dros point out that a recall test can show nothing retained while relearning is still faster, so a chart that labels the curve "percent remembered" is describing a different measure.
The 2015 replication
Murre and Dros (2015) repeated the experiment with one subject, one of the authors, who spent about 70 hours over 75 days learning 70 lists of nonsense syllables and relearning them after seven delays. Average savings fell from 42.1% at 20 minutes to 33.5% at one hour, 27.1% at nine hours and 23.9% at two days, and reached 9.0% at 31 days. The authors judged the replication a success.
The curve was not smooth. At one day, savings rose to 31.5%, above the nine-hour figure; the authors think the curve most probably jumps upwards at 24 hours, which some researchers attribute to sleep. Ebbinghaus had seen the same oddity in his own figures, found it not credible, and put it down to chance. The limits: one person, meaningless lists, relearning rather than recall.
How long should the gap between reviews be?
The best gap between reviews depends on how long the material has to last: about a fifth of the time to the test for a few weeks, falling to about a twentieth for a year. There is no single right interval, and Cepeda et al. (2008) put the practical consequence bluntly: to plan study, first decide how long you want to remember.
One detail changes how the table is read: in these experiments, the time before the test runs from the review to the test, not from first learning.
| Gap before the test (review to test) | Suggested review gap (best of those tested) | Source |
|---|---|---|
| 7 days | 1 day | Cepeda et al., 2008: 32 trivia facts, online volunteers |
| 10 days | 1 day | Cepeda et al., 2009, Experiment 1: 40 Swahili words, undergraduates |
| 35 days (5 weeks) | 11 days | Cepeda et al., 2008 |
| 70 days (10 weeks) | 21 days | Cepeda et al., 2008 |
| About 6 months | 1 month (28 days) | Cepeda et al., 2009, Experiment 2: facts and object names |
| 350 days | 21 days | Cepeda et al., 2008 |
| Several years | At least several months | Cepeda et al., 2008: the authors' advice, not a tested gap |
| Your own exam date | Worked out from the rows above | Each study's DOI resolves in EdCitation's Cite a source |
The gaps are the best of a short list that each experiment tried. In the 2008 study, the gaps for a 350-day test were 0, 1, 7, 21, 35, 70 and 105 days, so the true best could lie between 21 and 35.
What the 2009 experiments added
Cepeda et al. (2009) ran two laboratory studies with University of California, San Diego undergraduates. In the first, 182 students learned 40 Swahili-English word pairs, reviewed them after a gap of 0 to 14 days and were tested 10 days later. A one-day gap raised recall by 34% over a gap of a few minutes; stretching it to 14 days cost 11%.
In the second, 161 students learned little-known facts and the names of unfamiliar objects and were tested six months after the review. Here the one-day gap was far from best: a 28-day gap produced 151% more recall than no gap, and a one-day gap only 18% more.
Too long beats too short
In both papers, recall climbed steeply as the gap grew towards its best value and then fell away slowly. Cepeda et al. (2009) found that gaps longer than the best cost relatively little, not significantly though not trivially, and concluded that a too-short gap costs far more than a too-long one. For a schedule, that means: when unsure, wait longer.
The limits are the same in every row. The material was facts and word pairs, one review was tested rather than many, and the 2008 volunteers were online, not students in a course. No experiment here tested essays, mathematics or a whole module.
Should the gaps get longer each time?
Gaps that grow with each review are no better on average than equal gaps, according to a meta-analysis of the question.
Latimier et al. (2021) pooled 29 studies of spaced retrieval practice. Across 11 studies and 39 comparisons, spacing retrieval out beat massing it, with a large effect that stayed large after correcting for publication bias (g = 0.74). Across 16 studies and 54 comparisons of expanding against uniform schedules, the difference was close to zero (g = 0.034) and not significant.
When expanding gaps may help
The authors found one pattern: the more times an item was tested, the better the expanding schedule did against the uniform one. In their table that moderator reached p = .09, so it is a lead rather than a settled rule. Most studies were laboratory work with adults learning word pairs. The PDF notes a corrected publication in 2021; the record linked no correction notice, and we could not see what changed.
How does the Leitner system work?
The Leitner system is a set of flashcard boxes, each reviewed at a different interval: a card you answer correctly moves to a box reviewed less often, and a card you miss moves back to one reviewed more often. We have not read Sebastian Leitner's 1972 book, which is in German; the description here is Settles and Meeder's (2016).
In the variant they illustrate, the boxes stand for gaps of 1, 2, 4, 8 and 16 days, and every card starts in the one-day box. Unlike a fixed schedule, it adapts: an easy word soon moves to a rarely reviewed box while a hard one keeps coming back.
Setting up a paper Leitner box
These steps follow that description; the gaps are the illustrated variant, not a tested optimum.
- Write each fact as a question on one side of a card, the answer on the other.
- Put every new card in box 1, reviewed daily.
- Answer before you turn the card over; a glance at the answer is not a review.
- Move a card you got right to the next box (every 2, 4, 8, then 16 days); move a card you missed back to box 1.
- Keep one large deck rather than several small stacks. Kornell's (2009) abstract reports that one large stack of flashcards beat four smaller stacks studied separately, because small stacks shorten the gaps; spacing helped 90% of participants, yet after the first session 72% thought massing had worked better.
- Add cards for an essay's key sources too: build each reference from its DOI in EdCitation's Cite a source, and let the card hold what you learned from it.
Flashcard apps
Settles and Meeder, writing from Duolingo, report that it used a Leitner-like variant at launch, and that a model they trained on its data predicted recall with at least 45% less error than Leitner.
Anki Manual (n.d.) describes two schedulers: a legacy algorithm based on SuperMemo 2 and an alternative called FSRS. Under FSRS the learner sets a desired retention, 90% by default; higher settings shorten the intervals, above 90% the workload climbs very quickly, and in the manual's own example five more percentage points of retention needed study 35% more often. The manual also warns that pressing "Hard" on a card you actually forgot makes every later interval too long. We have not tested any app, and no study cited here compared them.
How do you build a revision calendar?
A revision calendar based on these findings sets a first review soon after learning, then returns at gaps sized to the time before the exam. The plan is ours, drawn from the studies above; no study tested it, and they used facts, not whole courses.
- Write down the exam date and the topics. Count the days from each topic's first lesson to the exam.
- Test yourself on each topic the day after you learn it. A one-day gap was best for a 10-day test (Cepeda et al., 2009), and a quick first check also shows what did not go in.
- Set the next gaps at about a tenth to a fifth of the time from learning to the exam. For an exam eight weeks away, that is roughly one to one and a half weeks between reviews.
- Keep the gaps equal unless you review often. Latimier et al. (2021) found no average advantage for expanding gaps.
- Round up, not down. A review a few days late costs far less than one too early.
- Review older topics alongside new ones. Each week, test yourself on this week's topic and one or two whose review falls due.
- Make the last review a few days before the exam, and give it to the material your self-tests showed was weakest.
An example for an exam eight weeks away
Suppose a topic is taught on day 1 and the exam is on day 56. Our calendar reviews it on day 2, day 12, day 23, day 34, day 45 and day 52: about eleven days apart, a fifth of the eight weeks, with a closing check four days before the exam. A topic taught in week 6 gets a review the next day and one before the exam.
How can EdCitation help you find and check the spacing research?
For tracing a claim about memory back to the paper that made it, EdCitation is the tool to use: each reference it gives you comes from the publisher's record, looked up rather than generated, which a chatbot cannot promise. It never writes any part of a student's work.
Searching for the spacing studies
Find sources covers about 300 million published works and takes a topic or the claim you need support for. A search in September 2026 for "spacing effect optimal interval between study sessions retention", sorted by citations, returned 11,933,324 works, and the list came back in descending order of citations. First came Cepeda and colleagues' 2008 ridgeline paper, cited 466 times, then Toppino and Cohen's 2009 paper on the testing effect and the retention interval (120). Two of the first ten were engineering papers, on pin fins and on wind turbines, so add your subject's own words.
Building the references
Cite a source turns a DOI, title, PubMed ID, ISBN or web address into an entry in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver, with a retraction screen. From the DOI 10.1371/journal.pone.0120644 it returned, in a rerun on 27 September 2026:
Murre, J. M. J., & Dros, J. (2015). Replication and analysis of Ebbinghaus’ forgetting curve. PLOS ONE, 10(7), Article e0120644. https://doi.org/10.1371/journal.pone.0120644
The publisher deposited the title in title case, and Cite a source set it in sentence case, as APA 7 wants for article titles; the entries for Cepeda et al. (2009), Latimier et al. (2021), Kang (2016) and Settles and Meeder (2016) came back in sentence case too. Our guide to title case or sentence case sets out the rule. None of the six sources checked was flagged as retracted.
Checking the finished list
Verify references takes a reference list or a whole paper, compares each entry with the publishers' records, marks any retraction, and keeps "could not check" apart from "not found". All nine references below came back verified. The seven entries with DOIs each came back with "The DOI resolves to this record and the title matches."; the Ebbinghaus translation with "Title, year and first author match a published record."; and the Anki manual page with "The page is at this address, and its title matches."
Free, Pro and Max
Find sources, Cite a source and Verify references are free, with no account. Pro, at $8 a month, adds References from a file, which checks every reference in a paper and matches every in-text citation to the list; see Pricing.
Quick questions
What is the best spaced repetition schedule?
No single spaced repetition schedule is best, because the right gap grows with the time to the test. For a test 10 weeks after the review, a 21-day gap did best of those Cepeda and colleagues tried; for six months, a 28-day gap.
Is the forgetting curve real?
The forgetting curve has been replicated: Murre and Dros (2015) repeated Ebbinghaus's experiment and found a similar curve, with a rise at 24 hours. Both studies used one person and nonsense syllables, so the exact percentages do not transfer to course material.
Do the gaps between reviews have to get longer each time?
Growing gaps do not have to be used: across 54 comparisons, Latimier and colleagues found no average difference between expanding and equal gaps. Expanding gaps may help when an item is tested many times.
Does the Leitner system work?
The Leitner system applies spacing and self-testing, both well supported, but no study we read tested the paper box itself. Settles and Meeder describe it and report that their trained model predicted recall better on Duolingo's data.
How do I check the studies on spaced repetition myself?
Search EdCitation's Find sources for the finding and the outcome, for example "spacing effect retention interval", then paste your list into Verify references to confirm each study exists and is not retracted.
References
- Anki Manual. (n.d.). Deck options. Retrieved September 27, 2026, from https://docs.ankiweb.net/deck-options.html
- Cepeda, N. J., Coburn, N., Rohrer, D., Wixted, J. T., Mozer, M. C., & Pashler, H. (2009). Optimizing distributed practice: Theoretical analysis and practical implications. Experimental Psychology, 56(4), 236–246. https://doi.org/10.1027/1618-3169.56.4.236
- 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. https://doi.org/10.1111/j.1467-9280.2008.02209.x
- Ebbinghaus, H. (1913). Memory: A contribution to experimental psychology (H. A. Ruger & C. E. Bussenius, Trans.). Teachers College, Columbia University. https://psychclassics.yorku.ca/Ebbinghaus/index.htm (Original work published 1885)
- Kang, S. H. K. (2016). Spaced repetition promotes efficient and effective learning: Policy implications for instruction. Policy Insights from the Behavioral and Brain Sciences, 3(1), 12–19. https://doi.org/10.1177/2372732215624708
- Kornell, N. (2009). Optimising learning using flashcards: Spacing is more effective than cramming. Applied Cognitive Psychology, 23(9), 1297–1317. https://doi.org/10.1002/acp.1537
- Latimier, A., Peyre, H., & Ramus, F. (2021). A meta-analytic review of the benefit of spacing out retrieval practice episodes on retention. Educational Psychology Review, 33(3), 959–987. https://doi.org/10.1007/s10648-020-09572-8
- Murre, J. M. J., & Dros, J. (2015). Replication and analysis of Ebbinghaus' forgetting curve. PLOS ONE, 10(7), Article e0120644. https://doi.org/10.1371/journal.pone.0120644
- Settles, B., & Meeder, B. (2016). A trainable spaced repetition model for language learning. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 1848–1858). Association for Computational Linguistics. https://doi.org/10.18653/v1/P16-1174