Spaced repetition is a calendar trick, not a memory hack. You front-load three short reviews in the first 48 hours, then space them out as accuracy stabilises. Skip the "magic interval" debate — pick one schedule, run it for two weeks, then adjust based on your own retention curve.
Most students hear about spaced repetition and immediately go shopping for an app. Two weeks later they have 1,400 cards on Anki, a half-built Quizlet deck, and a vague sense that everyone except them must be doing this right. The technique is real, the research is solid, and almost none of the gains come from the software. They come from putting reviews on a calendar and showing up when the calendar says to.
This is a working playbook — not a literature review. The science backs everything below, but the version you actually run on a Tuesday night matters more than the version that sounds clever in a journal.
What "spaced repetition" actually means
The cleanest definition comes from cognitive psychology: you review material at expanding intervals, and each successful retrieval strengthens the memory trace more than the previous one. Ebbinghaus mapped the forgetting curve in 1885 and noticed that the first re-exposure does the heavy lifting — without it, you lose around 60% of new content inside 24 hours.
What's changed since 1885 is the calibration. Cepeda and colleagues (2008) ran the spacing equation across 1,354 participants and found that the optimal gap between reviews is about 10-20% of the total time you need to remember something. If you need a fact for a test in 30 days, the next review should land 3-6 days out. For a test in a week, it's roughly 24 hours. This is the only number worth memorising from the research.
Three schedules that actually work
Pick one. Running all three in parallel is what made your last deck unmanageable.
The 20-minute daily
For when you have a steady drip of new material and zero patience for system maintenance.
| Day | Action | Time | |---|---|---| | Day 0 (lecture) | Take notes, mark 6-10 facts as "needs to stick" | during class | | Day 0 (evening) | Convert those facts into recall prompts | 8 min | | Day 1 | Retrieve from memory, mark hits/misses | 6 min | | Day 3 | Re-test only the misses + 50% random of yesterday's hits | 6 min | | Day 7 | Full deck cumulative | 10 min | | Day 21 | Spot-check, retire mastered cards to "monthly" | 5 min |
You're not chasing 100% retention. You're chasing "can recall under exam conditions in 4 weeks." Hits that show up four sessions in a row graduate out of the daily review and into a monthly bin.
The weekly cumulative
For dense content blocks — pharmacology, biochemistry, a whole semester of anatomy — where the volume is too high for daily card-level scheduling.
You break each week into a 4-day learning sprint and a 1-day pure-review day:
- Days 1-4: New material only. Two passes per topic: lecture notes → recall prompts.
- Day 5: 90 minutes of cumulative recall. No new material. You hit everything from this week and 30% sampled from last week. Failures get re-queued for next Day 5.
- Days 6-7: Active rest. One 20-minute review block max.
The Day-5 review is where retention actually moves. Most weeks people skip it for "one more lecture" — that's the choice that breaks the system.
The exam-block taper
Six weeks out from a cumulative exam (USMLE Step 1, Bacalaureat, finals week), you switch from new-material-with-review to review-with-occasional-new.
- Weeks 6-5: 70% new / 30% review.
- Weeks 4-3: 50% / 50%.
- Weeks 2-1: 20% new / 80% review.
- Last 72 hours: Zero new content. Two full cumulative passes, second one timed.
The taper is the single change that separates students who walk into the exam with the right facts cached versus students who learned an extra chapter the day before and forgot something they already knew.
What the research is and isn't saying
Three findings to internalise — and one that gets cited everywhere but doesn't translate to a real study schedule.
Retrieval beats review. Roediger and Karpicke (2006) showed that testing yourself produces ~50% more long-term retention than re-reading the same material for the same time. This is the cleanest result in cognitive psychology applied to studying, and it's the reason your prompts should be questions you have to answer out loud, not paragraphs you re-read in your head.
Spacing beats massing. Cepeda et al. (2006, meta-analysis of 254 studies, n>14,000) found massed practice — reading something three times in a row — gives a short-term boost that decays inside a week. Spacing the same three exposures over days produces 10-30% better retention at 1-month checkups. The trade-off is that spacing feels harder in the moment, which is why people stop doing it.
Desirable difficulty matters. Bjork & Bjork (2011) coined the phrase for the counterintuitive result that struggle during retrieval is when learning happens. If your prompts are so easy you nail them every time, the schedule is too tight. If you miss more than 20%, it's too loose. The sweet spot for working students is somewhere around 80-85% recall accuracy per session.
The thing that gets over-cited: "Reviews should be at 1 day, 7 days, 16 days, 35 days." Those numbers come from a single 1972 study by Pimsleur on language learning and have been quoted as universal ever since. They're a reasonable starting point. They're not the right answer for everyone. Calibrate based on your own retention.
Picking an interval algorithm — without overthinking it
Two main options if you use software:
- SM-2 (Anki's default): simple, conservative, multiplies intervals by a factor based on how well you recalled. Tends to over-schedule reviews. Good for high-stakes, low-tolerance-for-failure subjects (medicine, pilot exams).
- FSRS (Free Spaced Repetition Scheduler): a Bayesian model that learns your forgetting curve from your hit/miss history. Schedules fewer reviews for the same retention target. Anki has had it built-in since v23.10. Use this unless you have a reason not to.
Neither algorithm fixes a bad deck. A card with a vague prompt and a 200-word answer will fail under any schedule.
What "good cards" actually look like
The Karpicke & Blunt (2011) study compared concept-mapping (passive) against retrieval practice (active) and found a 50% retention advantage for retrieval — but only if the retrieval prompts were specific enough to require an answer rather than a recognition pattern.
Three rules:
- One fact per card. "Hyperkalemia causes peaked T waves and prolongs PR" is two cards.
- The prompt is a question, not a topic. "What ECG changes does hyperkalemia produce, in what order?" not "Hyperkalemia."
- Include the trigger. If you'll only need to recall it in the context of a specific patient presentation, put the presentation in the prompt: "70 y/o with renal failure presents with weakness — what ECG finding do you look for first?"
This is where most people's decks fail. They write cards that test recognition ("which of these is true about X?") and then wonder why they freeze when an exam asks for free recall.
Common ways the system breaks
You stop reviewing the day before a deadline. This is fatal. You skip Day 5, you skip the next Day 5, and by week 4 the cumulative load is unmanageable so you scrap the system. Solution: protect the review block. If a Day 5 is going to get crushed by an assignment, halve it to 45 minutes — don't skip it.
Your decks balloon. Going from 50 active cards to 800 in a month means you optimised for adding content over reviewing it. Solution: a hard cap on new cards per day (15-20 for most students; 30 if you're full-time studying for a board exam).
You confuse "I've seen this card a lot" with "I know this material." This is the most common failure mode. Recognition is not recall. Cover the answer, say it out loud, then reveal. If you can't say it out loud, mark it wrong even if it feels familiar.
You let the schedule slip during exam week. This is when spaced repetition pays off the most and gets discarded the most. The taper schedule above exists exactly for this — front-load the discomfort to the weeks where you can afford it.
What to do tonight
Pick one piece of material — a lecture, a chapter, an old exam question. Convert 8 facts into prompts using the three rules above. Tomorrow at the same time, sit down with the same 8 prompts and try to answer them out loud before peeking. That's day one of a system; everything else is calibration.
If you want the software to do the schedule maths for you, Anki with FSRS is the working default. If you want something that builds the cards from a PDF and runs the schedule for you, that's what Queazy does.
FAQ
Is spaced repetition useful for conceptual subjects, or just facts?
Both — but the prompt structure changes. For facts (anatomy landmarks, drug doses), you're testing direct recall. For concepts (why a mechanism produces a specific outcome), the prompt is a short scenario and the answer is an explanation in your own words. The Chi et al. (1994) work on self-explanation shows that producing the explanation under retrieval pressure is where conceptual understanding consolidates.
How long should each card take to review?
Under 10 seconds for the prompt, under 30 seconds for the answer if you have to think about it. Cards that take longer are usually two cards stitched together or have a prompt that's actually a topic.
What if I miss a week?
Resume at the next Day 5, do a cumulative review of the missed material, expect the recall accuracy to be 20-30% lower than your steady state, and don't panic. The schedule self-corrects within two weeks.
Anki, Quizlet, or Queazy?
Anki if you want maximum control and you're willing to maintain it. Quizlet if you need community decks and a friendlier UX. Queazy if you want it to read your PDFs and build the decks for you. None of them rank higher than a well-built deck — pick the one you'll actually open.
Sources
- Ebbinghaus, H. (1885). Über das Gedächtnis. English edition: https://psychclassics.yorku.ca/Ebbinghaus/
- 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.1467-9280.2008.02079.x
- 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. https://doi.org/10.1037/0033-2909.132.3.354
- Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249-255. https://doi.org/10.1111/j.1467-9280.2006.01693.x
- Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science, 331(6018), 772-775. https://doi.org/10.1126/science.1199327
- Bjork, R. A., & Bjork, E. L. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In Psychology and the Real World. https://bjorklab.psych.ucla.edu/wp-content/uploads/sites/13/2016/07/BjorkBjork2011.pdf
- Chi, M. T. H., de Leeuw, N., Chiu, M.-H., & LaVancher, C. (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18(3), 439-477. https://doi.org/10.1207/s15516709cog1803_3