Spaced repetition: what it is and how to use it

Spaced repetition has you review a fact shortly before you would forget it. Here is why it works, how to start, and the mistakes that make people quit.

5 min read

Spaced repetition is a way of reviewing in which each review is scheduled for the moment before you would forget. If you remember, the next review moves further away. If you forget, the item comes back soon. Over weeks, what you know needs fewer and fewer reviews, and what you keep missing gets more of your time.

#The method in four lines

1. Write what you want to remember as a question with one clear answer.
2. Try to recall the answer before you look at it.
3. Tell the system how it went, on a scale from "forgot it" to "knew it at once".
4. The system picks the next date. Sooner after a miss, later after a success.

Any set of paper cards and boxes can do this. Software does the bookkeeping for thousands of cards at once.

#Why it works

#Forgetting is steep at first, then slow

In 1885 Hermann Ebbinghaus published Über das Gedächtnis (in English, Memory: A Contribution to Experimental Psychology, 1913 translation). He learned lists of nonsense syllables, tested himself after delays, and measured how much less effort it took to learn a list again. Most of what he lost, he lost soon after learning, and the loss then flattened out. In 2015 Murre and Dros repeated the experiment with delays from 20 minutes to 31 days and got results similar to his original data.

Modern models describe the same shape. The chance of recalling something falls with time, and each successful review makes the fall slower.

The diagram is a sketch, not a measurement of anyone's memory. It shows the idea: a review is due when the chance of recall reaches your target, and every review stretches the time until the next one.

#Spacing beats cramming

In 2006 Cepeda and colleagues reviewed studies of distributed practice in verbal recall: 839 assessments in 317 experiments from 184 articles. The gap between study sessions and the time you need to remember work together: the longer you want to remember something, the longer the best gap.

#Recalling beats rereading

In 2006 Roediger and Karpicke had students read prose passages, then either take a recall test or reread. Five minutes later, rereading did better. Two days and a week later, the students who had been tested remembered more. The students who reread felt more confident about what they knew.

In 2013 Dunlosky and colleagues rated ten study techniques. Practice testing and distributed practice got the highest utility rating. Summarizing, highlighting and rereading got the lowest.

Spaced repetition software puts both findings in one habit: every review is a recall test, and the schedule spaces them.

#What decides the date

Anki's classic scheduler derives from SuperMemo 2, an algorithm by Piotr Wozniak. A newer open-source scheduler, FSRS, was developed by Jarrett Ye and contributors and has been built into Anki since version 23.10. FSRS tracks three things for each card:

• Retrievability: the chance you can recall it right now.
• Stability: how many days it takes for that chance to fall from 100% to 90%.
• Difficulty: how hard the material is, which sets how fast stability grows.

You choose a desired retention, the chance of recall you want on the day a card comes back. The FSRS documentation calls values between 70% and 97% reasonable. Higher means more reviews.

#How to start, step by step

Take an example: you travel to Spain in three months and want 150 words and phrases.

After a few weeks, check how many reviews you do per day. If the number only grows, add fewer new cards or lower the retention target.

#Mistakes that make people quit

• Starting with too many new cards. The first week feels light, the second week is a wall. Start low and raise it.
• Cards that are too big. A card holding a list of eight items is rarely answered cleanly. Split it.
• Learning what you do not understand. Wozniak's first rule in his twenty rules of formulating knowledge is not to learn what you do not understand. Memorizing a fact you cannot place keeps it fragile.
• Skipping days, then clearing the pile in one sitting. After a break, stop adding new cards until the reviews are back to normal.
• Rating everything Easy to finish faster. The schedule trusts your rating, so cards come back too late.
• Using cards for skills. Solving equations or writing an essay needs practice on real problems. Cards are for the facts those skills rest on.

#In Prior

Prior's Flashcards space schedules cards with FSRS-5 and the published default parameters. A card's schedule is computed from its review history, so two devices that reviewed offline end up with the same schedule once they sync.

• Create a deck, then add cards one by one, from a course, or from Anki. See Decks and packs and Import and export.
• Press the review button on the Flashcards home page (for example Review 23 cards). Space or Enter shows the answer, and 1 to 4 rate it: Again, Hard, Good, Easy. Review cards has the details.
• In Deck settings, New cards per day (20 by default), Maximum reviews per day (200) and Target retention (90%) are yours to change.
• How scheduling works explains what each rating does, and Flashcards overview shows the whole space.