# Spaced Repetition for USMLE: Anki & FSRS Guide

> How do you use spaced repetition for USMLE flashcards?

How spacing and retrieval work, which Anki and FSRS settings to start with, how to turn missed questions into cards, and how to size the daily load and clear a backlog.

Spaced repetition brings each flashcard back just before you are likely to forget it. For USMLE, write one fact per card, answer from memory before flipping, press Again whenever you forgot, and let FSRS set the intervals. Cap new cards by what they cost: Anki's manual estimates 20 new cards a day means about 200 reviews a day.

## How do you use spaced repetition for USMLE flashcards?

Spaced repetition pays off for USMLE when six habits run every day.

1. **Card the gaps.** Write a card, in your own words, when a lecture, question or self-test exposes a fact you could not produce.
2. **Keep one fact per card,** so it has one right answer and can be graded in seconds.
3. **Recall before you reveal.** Say or type the answer first; Anki's manual suggests showing it if nothing comes within about 10 seconds.
4. **Grade strictly.** Again if you could not produce it, Hard if correct but slow or doubtful, Good if correct with effort, Easy only if effortless.
5. **Review every day,** because skipped days stack into a backlog.
6. **Cap new cards** by the reviews they will create, worked out below.

## The science: forgetting curve, spacing effect, testing effect

Three findings from memory research explain why spaced recall beats rereading.

- **Forgetting curve.** Memories fade without review. A 2015 replication of Ebbinghaus's curve, with one subject relearning lists after 20 minutes to 31 days, produced results similar to his original data; the curve was not completely smooth and most probably jumps upward from the 24-hour point ([Murre & Dros, 2015](https://pubmed.ncbi.nlm.nih.gov/26148023/)).
- **Spacing effect.** Across 839 assessments in 317 experiments, the study gap giving the best retention grew as the time to the final test grew ([Cepeda et al., 2006](https://pubmed.ncbi.nlm.nih.gov/16719566/)). In more than 1,350 people, the best gap was about 20 to 40% of a 1-week test delay but 5 to 10% of a 1-year delay ([Cepeda et al., 2008](https://pubmed.ncbi.nlm.nih.gov/19076480/)), so a later test needs a longer gap.
- **Testing effect.** Restudying a passage won on a test 5 minutes later, but recall tests produced substantially greater retention at 2 days and 1 week, even though restudying raised students' confidence ([Roediger & Karpicke, 2006](https://pubmed.ncbi.nlm.nih.gov/16507066/)). In pediatric and emergency medicine residents, repeated testing with feedback beat repeated study, 39% to 26%, about 6 months later ([Larsen et al., 2009](https://pubmed.ncbi.nlm.nih.gov/19930508/)).

A review of 10 study techniques rated practice testing and distributed practice high utility, and rereading and highlighting low utility ([Dunlosky et al., 2013](https://pubmed.ncbi.nlm.nih.gov/26173288/)).

| Method | Evidence | Best USMLE use |
| --- | --- | --- |
| Rereading or highlighting | Low utility; rereading beat testing only at a 5-minute test | First pass on a topic you do not yet understand |
| Self-testing in one sitting | Beat restudying at 2 days and 1 week | Same-day check after a lecture |
| Spaced repetition | Effect size 0.78 over standard study in 13 medical-education studies | Facts that must last months |
| Spaced case questions | Better end-of-year retention in a randomized trial | Applying facts in exam format |

## What studies of medical students show about Anki and exam scores

Most Anki studies in medical school are observational: they show associations with scores, not proof that the cards caused them.

- **[Deng et al., 2015](https://pubmed.ncbi.nlm.nih.gov/26498443/) (72 students).** Boards-style questions completed and unique Anki cards seen were independent predictors of numeric Step 1 score; each additional 445 questions, or 1,700 unique Anki cards, was associated with one more point after controlling for other academic and psychological factors.
- **[Wothe et al., 2023](https://pubmed.ncbi.nlm.nih.gov/37187920/) (165 students).** Daily Anki use correlated with a higher Step 1 score but not with [Step 2 scores](/usmle/step-2-ck-score-percentiles).
- **[Spence et al., 2025](https://pubmed.ncbi.nlm.nih.gov/41939075/) (176 first-years).** Anki use (82% of respondents) was not associated with better in-house exam performance; higher wellness and at least seven hours of sleep went with better written-exam scores.
- **[Frappa et al., 2026](https://pubmed.ncbi.nlm.nih.gov/42183420/) (systematic review, 11 studies).** Three studies showed a consistent positive link between regular Anki use and Step 1 performance, with high-frequency users 4 to 13 points ahead of minimal users; course exams were mixed, and the one Step 2 CK study found no significant benefit.
- **[Maye & Hurley, 2026](https://pubmed.ncbi.nlm.nih.gov/41601436/) (meta-analysis, 21,415 learners).** Spaced repetition beat standard study (standardized mean difference 0.78), pooling faculty-made or third-party flashcards, emailed or CME-delivered questions and spaced classroom quizzes.

Those Step 1 scores predate [pass/fail reporting](/usmle/step-1-passing-score), which applies to exams taken on or after 26 January 2022 ([USMLE](https://www.usmle.org/usmle-step-1-transition-passfail-only-score-reporting)), and no study shows that a card count guarantees a result.

## SM-2 vs FSRS: how Anki sets each interval

Anki offers two schedulers: the legacy SM-2 algorithm and FSRS, the Free Spaced Repetition Scheduler, supported from Anki 23.10.

**SM-2** multiplies the last interval by a per-card ease that starts at 2.50, so Good on a card last seen 10 days ago brings it back in about 25 days, and later grades push the ease up or down. Pressing Again on a review card is a lapse: by default the interval resets and the card returns the day after relearning.

**FSRS** estimates each card's difficulty, stability and retrievability, fits its parameters to your review history, and schedules each review for when predicted recall falls to your desired retention.

**Desired retention** is the most important FSRS setting. At the default 90%, you should recall about 9 cards in 10 when they come due; raising it shortens intervals and adds reviews, and above 90% the workload rises very quickly. Expect Again on about 5 to 20% of cards, the range the manual calls typical.

## Anki settings for USMLE: starting points, not rules

Anki's manual advises a few weeks on the defaults before changing options, so treat these values as starting points to test.

| Setting | What it changes | Starting point | Manual guidance |
| --- | --- | --- | --- |
| New cards/day | Unseen cards added daily | From your review budget, below | 20 a day means roughly 200 reviews a day |
| Maximum reviews/day | Daily review ceiling | High enough to rarely bind | Smooths peaks, as after a week off |
| New cards ignore review limit | New cards once the cap is hit | Off, the default | Stop new cards while behind |
| Learning steps | Same-day repeats of new cards | One or two short steps, such as 10m | With FSRS, few steps, all under a day |
| Relearning steps | Same-day repeat after a lapse | One step, such as 10m | Avoid steps of a day or more with FSRS |
| Desired retention | Target recall at review | 90%, the default | Workload climbs fast above 90%; stay under 97% |
| FSRS parameters | Your forgetting model | Optimize after a few hundred reviews, then about monthly | Never copy someone else's |
| Leech threshold | Lapses before Anki tags a leech | 8, the default | Rewrite, split or delete leeches |
| Review sort order | Which due cards come first | Default due-date order | Relative overdueness or ascending retrievability for a big backlog |

## How to write flashcards that survive a vignette

Anki's manual points card writers to Piotr Wozniak's 20 rules of formulating knowledge, which begin with understanding before memorizing and include the minimum information principle.

- **Understand first.** A mechanism you cannot explain tends to become a leech.
- **One fact, one answer.** "Which cause of nephrotic syndrome shows spike-and-dome subepithelial deposits?" (membranous nephropathy) is a card; "describe the nephrotic syndromes" is a study session.
- **Split lists.** Wozniak's rules call sets costly and suggest cloze deletion for enumerations, one deletion per item.
- **Cloze the tested term,** never a filler word; `{{c1::pyridoxine::vitamin}}` shows "vitamin" as a hint.
- **Occlude images.** Anki 23.10 and later build in image occlusion: Hide All, Guess One masks every label; Hide One, Guess One leaves the rest as context.
- **Separate look-alikes.** If two facts keep colliding, the manual's leech chapter suggests suspending one until the other is solid.
- **Keep the source.** Put the missed question's clue, in your own words, in the cloze Extra field, shown on the answer side.

| Card type | Use it for | Example |
| --- | --- | --- |
| Basic | Discriminators and "why" questions | Which antituberculosis drug causes optic neuropathy with loss of acuity and color vision? -> Ethambutol |
| Cloze | A fact in its context | Isoniazid causes peripheral neuropathy by depleting `{{c1::pyridoxine}}` |
| Two-deletion cloze | Cause-and-effect chains | Warfarin inhibits `{{c1::vitamin K epoxide reductase}}`, cutting synthesis of factors `{{c2::II, VII, IX and X}}` |
| Image occlusion | Anatomy, histology, pathways, imaging | A brachial plexus diagram with one nerve hidden per card |

A missed question shows how the rules combine: a patient on tuberculosis treatment develops tingling in both feet, you chose ethambutol, and the answer is isoniazid. Make the isoniazid cloze for the gap and the ethambutol card for the discriminator, since the real error was filing ethambutol under the wrong toxicity. The [high-yield drug side effects](/blog/high-yield-drug-side-effects-usmle) guide and the [Pharmacology](/topics/pharmacology) hub collect more drug facts worth carding.

## Premade deck or your own cards?

Premade decks and your own cards do different jobs: one supplies breadth, the other fixes the errors costing you points.

A large community deck such as AnKing covers material you have not met yet; one approach: keep it suspended and unsuspend each system as you study it. Writing your own has a real cost. When one medical school's pharmacology faculty built a 1,007-card Anki deck, each card took about 7 minutes, and student users scored lower than non-users, a gap that lost significance after adjusting for self-rated academic performance ([Harder et al., 2026](https://pubmed.ncbi.nlm.nih.gov/42820988/)). Inside a dedicated period, skip big imports and card only what your questions expose.

## How many new cards per day? Plan the review load first

Your new-card limit, not your deck size, sets tomorrow's workload, because each new card keeps returning for weeks.

Anki's manual estimates that a steady 20 new cards a day produces roughly 200 reviews a day (a ten-to-one rule of thumb, not a fixed ratio), and warns that users who added hundreds in their first days were soon overwhelmed. Work backward from your time:

1. **Measure your speed:** divide the daily study time Anki shows by the cards you reviewed.
2. **Find your review budget:** minutes x 60 / seconds per review, so 60 minutes at 10 seconds allows 360 reviews.
3. **Set the cap near a tenth of that,** about 36 here, then start nearer 25 to leave room for the new cards themselves and bad days.
4. **Recheck after two weeks,** lowering the cap rather than skipping days.

Where that block sits in your day belongs to your [Step 1 study schedule](/blog/usmle-step-1-study-schedule).

## Worked example: one week of new cards and reviews

A student eight weeks before dedicated has an older deck showing about 150 reviews a day and adds 30 new cards a day, Monday to Saturday, keeping Sunday for reviews. For visible arithmetic, assume each new card is passed and returns 1 day and 4 days after it is learned, reviews take 10 seconds, and a new card takes 20 seconds including its learning step.

| Day | New cards | Older-deck reviews | Reviews of this week's cards | Total reviews | Minutes |
| --- | --- | --- | --- | --- | --- |
| Monday | 30 | 150 | 0 | 150 | 35 |
| Tuesday | 30 | 150 | 30 | 180 | 40 |
| Wednesday | 30 | 150 | 30 | 180 | 40 |
| Thursday | 30 | 150 | 30 | 180 | 40 |
| Friday | 30 | 150 | 60 | 210 | 45 |
| Saturday | 30 | 150 | 60 | 210 | 45 |
| Sunday | 0 | 150 | 60 | 210 | 35 |
| Week | 180 | 1,050 | 270 | 1,320 | 280 |

Friday's 60 are Thursday's cards at 1 day plus Monday's at 4 days. The week costs 4 hours 40 minutes, but the 180 new cards keep returning: by the ten-to-one rule of thumb, 30 a day kept up steadily would settle near 300 daily reviews, about 50 minutes, before any older cards. Relearning for the roughly 1 card in 10 forgotten at 90% retention comes on top.

## How to clear a flashcard backlog

A backlog clears fastest when new cards stop and every spare review goes to the overdue pile, as Anki's manual advises.

1. **Set new cards to zero;** adding them while behind makes the backlog worse.
2. **Cap each session** with maximum reviews/day at a size you will finish.
3. **Reorder a large backlog** by relative overdueness (SM-2) or ascending retrievability (FSRS), which put the cards you most likely forgot first.
4. **Never reset overdue cards.** Reset sends a card to the end of the new queue, while answering it late lets Anki factor in the delay, so you resume where you left off.
5. **Suspend or delete** obscure cards you will not need.

Example: with 1,200 overdue cards, 150 newly due each day and time for 300 reviews, 150 a day go to the pile, which clears in about 1,200 / 150 = 8 days.

## How flashcards and a question bank work together

A question bank finds gaps and trains application; flashcards keep each fix from fading.

Step 1 "assesses whether you understand and can apply important concepts of the sciences basic to the practice of medicine" ([USMLE](https://www.usmle.org/step-exams/step-1)), so cards are storage, not the test. Clear due cards early, run your blocks, then turn each miss into one or two cards of your own using the error log from [how to review UWorld incorrect questions](/blog/how-to-review-uworld-incorrect-questions). StepGenie's flashcards are premade: 20,000+ cards in Step 1 and Step 2 decks, scheduled with FSRS.

> **10-second flashcard check:** 1) Answer before you flip. 2) Forgot means Again; Hard is a pass. 3) One fact per card. 4) Rule of thumb: ten daily reviews per daily new card. 5) Behind? Zero new cards, no resets. 6) Questions find gaps; cards keep them closed.

## Flashcard traps that cost points on test day

Much wasted review time comes from habits that make your statistics look better than your memory.

- **Flipping before answering.** Recognition is not recall; the manual says a partial answer that would fail outside Anki fails inside it.
- **Pressing Hard on a forgotten card.** FSRS reads Hard as a pass and stretches the interval, the one habit the manual says FSRS cannot adapt to.
- **Trusting the feeling of knowing.** Once students could recall a word, more studying did nothing for delayed recall while more testing helped a lot, and their predictions of performance were uncorrelated with it ([Karpicke & Roediger, 2008](https://pubmed.ncbi.nlm.nih.gov/18276894/)).
- **One-way cards.** Vignettes often give the toxicity and ask for the drug, so add the reverse card.
- **Reviewing deck by deck.** The manual advises against showing each subdeck in turn, because a consistent order makes answers easier to guess from context and leads to weaker memories.

## Sources

- Memory research (PubMed): [Murre & Dros 2015](https://pubmed.ncbi.nlm.nih.gov/26148023/), [Cepeda 2006](https://pubmed.ncbi.nlm.nih.gov/16719566/), [Cepeda 2008](https://pubmed.ncbi.nlm.nih.gov/19076480/), [Roediger & Karpicke 2006](https://pubmed.ncbi.nlm.nih.gov/16507066/), [Karpicke & Roediger 2008](https://pubmed.ncbi.nlm.nih.gov/18276894/), [Larsen 2009](https://pubmed.ncbi.nlm.nih.gov/19930508/), [Dunlosky 2013](https://pubmed.ncbi.nlm.nih.gov/26173288/)
- Medical-student studies (PubMed): [Deng 2015](https://pubmed.ncbi.nlm.nih.gov/26498443/), [Wothe 2023](https://pubmed.ncbi.nlm.nih.gov/37187920/), [Spence 2025](https://pubmed.ncbi.nlm.nih.gov/41939075/), [Harder 2026](https://pubmed.ncbi.nlm.nih.gov/42820988/), [Frappa 2026 systematic review](https://pubmed.ncbi.nlm.nih.gov/42183420/), [Maye & Hurley 2026 meta-analysis](https://pubmed.ncbi.nlm.nih.gov/41601436/), [Kerfoot 2007 randomized trial](https://pubmed.ncbi.nlm.nih.gov/17209889/)
- USMLE: [Step 1 overview](https://www.usmle.org/step-exams/step-1), [Step 1 pass/fail transition](https://www.usmle.org/usmle-step-1-transition-passfail-only-score-reporting)
- Example pharmacology: [antitubercular drug toxicity (StatPearls)](https://pubmed.ncbi.nlm.nih.gov/30285383/), [ethambutol optic neuropathy review (PubMed)](https://pubmed.ncbi.nlm.nih.gov/36872667/)
- Anki's official manual (Deck Options, Studying, Adding/Editing, Leeches) and Piotr Wozniak's "Twenty rules of formulating knowledge" (1999), read 7 October 2026

Put high-yield facts on an FSRS schedule that brings each back before test day: [Try StepGenie spaced repetition flashcards](https://dashboard.stepgenie.app/sign-up).

## Frequently asked questions

### How many new Anki cards should I do per day for Step 1?

Pick the number from the review time you can protect, not from the size of your deck. Anki's manual estimates that a steady 20 new cards a day leads to roughly 200 reviews a day, which suggests a rule of thumb of about ten daily reviews per daily new card. With 60 minutes at about 10 seconds per review you can afford roughly 360 reviews, which points to a cap near 36; starting nearer 25 leaves room for the new cards themselves and for bad days.

### What desired retention should I use in FSRS?

Start at the default of 90%, which Anki's manual describes as a good balance of retention and workload: you should remember about 9 cards in 10 when they come due. Raising it shortens intervals and adds reviews, and the manual warns that above 90% the workload rises very quickly and that above 97% it can be overwhelming. Its experimental Help Me Decide simulator shows your own trade-off before you change the value.

### Does Anki improve Step 1 scores?

Several observational studies link regular Anki use to higher Step 1 scores, and a 2026 systematic review found high-frequency users 4 to 13 points ahead of minimal users. These are associations, not proof that the cards caused the difference, and they come from the three-digit era: Step 1 is reported pass/fail for exams taken on or after 26 January 2022. Results for in-house exams were mixed, and the one Step 2 CK study found no significant benefit.

### Should I use a premade deck like AnKing or make my own cards?

Use both, for different jobs. A large premade deck covers material you have not met yet, and unsuspending it topic by topic keeps new cards in step with your coursework. Your own cards should capture the specific facts your question blocks expose, written in your own words with one fact per card. Writing cards is slow: one faculty team averaged about 7 minutes per card when it built a 1,007-card pharmacology deck.

### How do I catch up on an Anki backlog?

Set new cards to zero until the overdue pile is gone, as Anki's manual recommends, and work through a daily number you can actually finish. For a large backlog, switch the review order to relative overdueness under SM-2 or ascending retrievability under FSRS, so the cards you are most likely to have forgotten come first. Do not reset overdue cards: Anki factors in the delay when you answer them late, so you resume where you left off.

---

Canonical page: [Spaced Repetition for USMLE: Anki & FSRS Guide](https://www.stepgenie.app/blog/spaced-repetition-usmle-flashcards)
Topic hub: [Study Strategy](https://www.stepgenie.app/topics/pharmacology)
