Review less often to retain longer
Spaced repetition is a revision strategy that schedules each review at an optimal moment: just before memory strength declines significantly. The interval between reviews expands each time retrieval succeeds -- from one day, to three days, to a week, to a month, to several months, eventually to years.
The result is counterintuitive: reviewing less frequently -- but at precisely the right time -- produces stronger and longer-lasting retention than frequent unscheduled review. The spacing is not incidental; it is the mechanism.
The spacing effect: what 100 years of research shows
The spacing effect -- the finding that distributed practice outperforms massed practice for long-term retention -- is one of the oldest and most replicated findings in cognitive psychology. It was first described by Ebbinghaus in 1885, studied experimentally throughout the 20th century, and confirmed at scale in hundreds of modern studies.
Kang (2016), in a synthesis published in Policy Insights from the Behavioral and Brain Sciences, concludes that under most tested conditions, spaced practice outperforms massed practice at equal effort -- often by a factor of two or three at 30-day retention. Cepeda et al. (2006) analysed 317 experiments with 839 effect-size estimates: distributed practice improved memory in 96% of studies. Few findings in cognitive psychology are this consistent. The effect holds across age groups, content domains, and learning methods -- from vocabulary to medical facts to mathematical concepts.
A 2024 study on medical students using spaced-repetition flashcard apps reported 30-40% exam-score improvement versus a control group using traditional methods -- with comparable total study time. The gain came not from studying more, but from studying at better-timed intervals.
Usage of Spaced Repetition Flashcards in Medical Education, PMC 2024The link between spaced repetition and the forgetting curve
Spaced repetition is a direct response to the Ebbinghaus forgetting curve: without any review, newly learned information drops to roughly 30% retention within 24 hours and continues declining. The forgetting curve shows that this decay is not linear -- it is exponential, with the steepest drop occurring in the hours immediately after learning.
Each spaced review resets the curve -- but at a higher starting point, and with a shallower slope. The first review might bring retention back to 100% and extend the next forgetting window to 3 days. The second review extends it to a week. By the fifth or sixth well-timed review, the forgetting slope has flattened to the point where retention remains stable for months or years. This compounding dynamic is why spaced repetition feels slow to get started -- the first few weeks of review seem unimpressive -- but becomes dramatically more efficient over time as hard-won stable memories require progressively less attention.
How each review changes the curve
Ebbinghaus observed not only the forgetting curve but what he called the learning curve for relearning: information previously learned requires fewer repetitions to relearn than brand-new material. This "savings effect" is the direct predecessor of spaced repetition -- each review strengthens the underlying memory trace, making future forgetting slower.
Modern memory research formalises this as memory stability: a property of each individual memory trace that increases with each successful retrieval. A memory with high stability can persist for months without review. A memory with low stability (newly learned, or not reviewed in a long time) decays rapidly. Spaced repetition systems track this stability and schedule reviews to keep each trace above a target retention threshold.
Why spaced repetition works at the neurological level
Why does spacing work? Smolen, Zhang, and Byrne (2016) provide one of the clearest explanations in Nature Reviews Neuroscience. Memory consolidation requires cascading biochemical and structural changes at synapses over time. Reviewing too soon after initial learning adds little because these processes have not yet run their course. Reviewing after a meaningful delay -- when memory begins to weaken -- forces active reconstruction, which triggers the consolidation cascade again from a stronger baseline.
The result is a compounding effect: each well-timed review increases not just current retention but future stability. Memories reviewed with appropriate spacing are not just retrieved -- they are rebuilt, each time more durably.
Desirable difficulty: why harder review is better review
Robert Bjork (1994) formalised the concept of desirable difficulties: learning conditions that feel harder short-term can produce better long-term outcomes. Spacing is one of the most robust documented forms. When a review occurs after a long enough interval that retrieval requires genuine effort, the memory trace is strengthened more than if the same review had occurred earlier when retrieval was effortless.
This has a practical implication that many learners resist: if a flashcard review session feels easy -- if you are getting everything right without any struggle -- you are probably reviewing too frequently. The optimal interval is one where you succeed, but only with noticeable retrieval effort. A system that makes you feel like you are forgetting more than you would like may actually be calibrated correctly. The discomfort of near-forgetting, followed by successful retrieval and confirmation, is precisely the experience that produces durable memory. Comfortable review sessions that feel effortless are often the least productive.
Stickgold (2005) demonstrated that sleep plays an active role in memory consolidation -- neural replay during slow-wave and REM sleep strengthens recently encoded traces. Spaced repetition and sleep are synergistic: a review session before sleep gives consolidation machinery the maximum opportunity to process and strengthen what was recalled.
Stickgold (2005), Sleep-Dependent Memory Consolidation, Nature, 437What makes a review interval optimal
Not all intervals are equal. Reviewing too early -- when retention is still very high -- produces little consolidation benefit because the memory is already strong. Reviewing too late -- after retention has collapsed -- requires essentially relearning the material from scratch, as if the previous session never happened.
The optimal interval sits between these two extremes: long enough that genuine retrieval effort is required, short enough that the trace has not degraded below the threshold for successful recall. Research suggests this optimum typically corresponds to a retention level around 70-90% at the moment of review -- the "forgetting threshold." Targeting this zone consistently is what separates a well-calibrated SRS from both inefficient cramming (intervals too short) and forgetting-driven relearning (intervals too long).
The gap effect and expanding intervals
Cepeda et al. (2009) studied optimal study-to-test intervals across thousands of question-answer pairs and found that the ideal gap between review sessions grows as the total time to the test increases. For a test in one week, reviewing after one day is optimal. For a test in six months, reviewing after weeks-long gaps is optimal. The ratio of study gap to retention interval is roughly 10-20%.
This is the basis for the expanding-interval logic in SRS systems: the first interval might be 1 day, then 3 days, then 10 days, then 30 days. Each successful retrieval at these intervals signals sufficient memory stability to justify extending the next gap. The growing interval is not arbitrary -- it is calibrated to the forgetting dynamics of that specific memory in that specific learner. A card you consistently recall correctly sees its interval double or triple each time; a card you struggle with stays in short-interval territory until it stabilises. Over weeks and months, easy material migrates to very long intervals while difficult material receives the additional review time it needs.
How SRS algorithms compute intervals automatically
Manual spacing is possible but tedious: you would need to track optimal intervals for hundreds or thousands of individual cards, each with its own review history. SRS (Spaced Repetition System) algorithms automate this computation. After each review, you provide a difficulty or confidence rating, and the algorithm calculates the next optimal interval for that specific card.
SM-2: the foundational algorithm
SM-2, developed by Piotr Wozniak in the late 1980s for his SuperMemo software, is the first widely adopted SRS algorithm and the direct ancestor of most modern flashcard systems, including Anki. After each review, the user rates recall ease on a scale (typically 0-5). The algorithm uses this rating plus the card's previous ease factor to compute the next interval.
The interval formula in SM-2 is: I(n) = I(n-1) x EF, where EF (ease factor) starts at 2.5 and adjusts based on ratings. A card rated "easy" consistently sees its EF increase, extending future intervals; a card frequently rated "hard" keeps a lower EF and short intervals. SM-2 was a revolutionary contribution to learning science that made automated spaced repetition accessible to millions of learners, but it has known limitations: its parameters are fixed rather than learned from individual data, it cannot distinguish between stability and retrievability as separate memory properties, and its interval estimates drift from optimal over long time horizons -- particularly for material that has been studied intensively or for very long retention goals.
FSRS: the next-generation algorithm
FSRS (Free Spaced Repetition Scheduler), developed by Jarrett Ye and released in 2022, represents a significant advance over SM-2. Where SM-2 uses a single ease factor per card, FSRS models two separate memory properties: stability (how long the memory can persist before decaying below a target threshold) and retrievability (the current probability of successful recall given elapsed time since the last review).
FSRS fits these parameters to each learner's actual recall history using machine learning, rather than applying fixed formulas. This allows the algorithm to adapt to individual differences in forgetting rates, the difficulty of specific content types, and the learner's general retention level. Comparative analyses on large real-world datasets consistently show FSRS outperforming SM-2 on interval precision -- meaning fewer unnecessary reviews and fewer missed reviews, translating to better retention for the same total time spent. Memia uses FSRS as its scheduling engine, which means your review schedule becomes more accurate the more cards you review -- a compounding advantage that grows over time.
With a good SRS, you do not need to manage intervals manually -- the algorithm handles all scheduling. Your only role is to grade honestly after each card. If recall was uncertain or slow, do not rate it as easy. Scheduling quality is entirely determined by feedback quality: overrating easy cards extends intervals prematurely; underrating hard ones makes you review more than necessary.
Manual vs automated spaced repetition
SRS software is the most efficient implementation for most learners, but spaced repetition existed as a paper system long before computers. Understanding both approaches helps you choose the right tool for your situation.
The Leitner box system
Sebastian Leitner formalised the first manual SRS in the 1970s in his book So lernt man lernen. The system uses a set of physical boxes labelled with review frequencies: Box 1 is reviewed daily, Box 2 every three days, Box 3 every week, Box 4 every two weeks, Box 5 every month. Cards start in Box 1. When you answer correctly, the card moves to the next box. When you answer incorrectly, it returns to Box 1.
The Leitner system is a practical approximation of the spacing principle -- it does not compute personalised intervals, but it ensures that easy cards are reviewed less often and difficult cards more often. It works well for decks of up to a few hundred cards and requires no technology. For language learners, exam preparation with a limited scope, or anyone who prefers physical materials, the Leitner system is a solid implementation that meaningfully outperforms unstructured review.
When to switch from paper to software
The Leitner system becomes unwieldy at scale. Managing 500+ cards across multiple boxes, tracking overdue reviews, and handling the growing complexity of which box a card belongs to after multiple failed attempts creates significant organisational overhead. At that volume, the system starts requiring more effort to manage than to actually review.
Software SRS tools (Anki, Memia, and similar) handle this complexity automatically. They store the full review history of every card, compute the next interval in milliseconds after each rating, and surface exactly the cards due for review each day -- no manual tracking needed. For learners building a deck beyond a few hundred cards, or maintaining knowledge across multiple subjects simultaneously, software SRS is the practical choice.
If you are new to spaced repetition, starting with a paper Leitner system for 2-4 weeks can build intuition for how intervals work before committing to a software workflow. Once you understand the spacing logic and have identified what content you want to retain long-term, moving to a software SRS with per-card interval precision becomes straightforward.
What spaced repetition does not do
Spaced repetition is one of the most powerful tools in a learner's toolkit, but it has clear limits. Understanding them prevents misapplication and helps you combine it effectively with other strategies to build robust, applicable knowledge rather than just a strong test score.
Where spaced repetition falls short
Spaced repetition does not replace understanding. Memorising a definition you do not understand is possible, but the resulting knowledge is fragile -- it does not transfer to new problems, cannot be applied flexibly, and is difficult to connect to other knowledge. The best flashcard decks test comprehension, not just surface recall: "Why does X happen?" rather than "What is the definition of X?"
Spaced repetition is also not well-suited to procedural skills. Learning to code, play an instrument, perform a clinical procedure, or write in a foreign language requires real practice with feedback -- not just recall of propositions about how to do it. Active recall and spaced repetition support the declarative knowledge layer (knowing what to do and why), but they cannot substitute for deliberate practice of the skill itself. Knowing the rules of counterpoint does not make you a composer; knowing the definition of recursion does not make you a programmer. The SRS handles memory consolidation; skill development requires separate, targeted practice.
What works well alongside spaced repetition
Spaced repetition works best as a memory consolidation backbone, not as the sole learning strategy. It pairs naturally with deep-processing activities that create the material being memorised: conceptual explanations (Feynman method), problem-solving practice, interleaved study sessions, and elaborative interrogation (asking "why" after learning each fact).
A practical integration: use initial study (reading, attending a lecture, working through problems) to build understanding, then convert key concepts into flashcards and enter them into an SRS. The spaced repetition system maintains and consolidates what you understood during the initial study phase. Neither component works as well alone -- understanding without review fades quickly; memorising without understanding produces brittle knowledge that cannot be applied flexibly. A common pitfall is making too many cards too quickly during initial study, before the material has been properly understood. Better to make fewer, well-understood cards and review them faithfully than to flood the deck with half-understood material that becomes a chore to review.
How Memia implements spaced repetition
Memia is built around FSRS spaced repetition from the ground up. Every card you add enters the scheduling algorithm immediately, and after each review your confidence rating updates that card's stability and retrievability estimates. The next review date is computed in real time based on your personal forgetting curve for that specific card -- not a generic schedule applied uniformly to everyone.
Each review session surfaces only the cards due today, sorted by urgency. Cards you are close to forgetting appear first. Cards with high current stability are not shown until their optimal review window opens -- which means you never waste time reviewing what you already know solidly. This focus on what actually needs attention is what makes short daily sessions with Memia outperform longer but unfocused study sessions.
You can import your own content (text, PDF, structured lists) and let the AI generate a complete flashcard deck, or browse the public catalogue for pre-built decks on hundreds of topics. Multiple question formats -- multiple choice, true/false, and open-ended -- ensure that retrieval is genuinely active rather than recognition-based. Progress dashboards let you track retention by topic and identify which knowledge areas are stable vs which need more review time.
Memia handles all scheduling automatically. You only need to review and rate honestly -- the algorithm does the rest. Start with a free account and add your first cards today.
Frequently asked questions
How many cards should I review per day with spaced repetition?
It depends on how many new cards you add and your available study time. A sustainable starting point is 10-20 new cards per day, which typically generates 50-150 reviews per day as the deck matures. Adding too many new cards too fast creates a review backlog that compounds over weeks -- it is better to start conservatively and increase gradually.
What happens if I miss several days of review?
A backlog of overdue cards accumulates. It is usually better to recover progressively -- clearing part of the backlog each day -- than to attempt everything at once, which can be demotivating. Most SRS apps handle overdue cards by recalibrating intervals downward: a card that has been overdue for three weeks will get a shorter next interval than one reviewed on time, reflecting the likely retention drop.
Is spaced repetition effective for Asian languages like Japanese or Mandarin?
Yes -- it is one of the most impactful use cases for SRS. High-character-count writing systems like Japanese (kanji) and Mandarin (hanzi) require memorising thousands of distinct characters. Spaced repetition is the most efficient known method for building and maintaining this kind of large-volume declarative knowledge base.
Can I combine spaced repetition with other study methods?
Absolutely -- and you should. Spaced repetition is a memory consolidation tool, not a full learning system. It works best combined with deep understanding methods (Feynman explanations, worked examples, problem-solving) and real practice for any procedural skills. Think of SRS as the system that keeps what you learned from fading, while other methods build the understanding that makes learning worthwhile.
What is the difference between SM-2 and FSRS?
SM-2 uses a fixed ease-factor formula to compute intervals and does not adapt to individual learning patterns beyond the card's own history. FSRS uses a two-component model (stability + retrievability) and fits its parameters to your actual recall data, producing more accurate interval predictions especially over long time horizons. In practice, FSRS tends to schedule fewer unnecessary reviews while maintaining the same or better retention.
How is spaced repetition different from just reviewing regularly?
Regular review on a fixed schedule (e.g., every Sunday) treats all material the same regardless of how well it is retained. Spaced repetition schedules each individual card based on its own history: items you know well get long intervals, items you struggle with get short ones. This personalisation is what makes SRS dramatically more efficient than calendar-based review.
How long before I see results with spaced repetition?
Improved retention is measurable within a few weeks for material you are actively reviewing. The biggest gains appear over months, as the compounding effect of successive reviews pushes stable memories to very long intervals -- freeing daily review time for new material while maintaining existing knowledge. Most learners notice qualitatively better recall after 3-4 weeks of consistent daily review.
← Previous article: Active recall
Next article: The illusion of competence -- why rereading is not enough