There are a lot of things that techie types esteem beyond their worth. Spaced repetition systems (SRS) sound like unnecessary complication on top of flashcards, and thus like an obvious example of that phenomenon. They are not.

Here is the best simple explanation of spaced repetition I’d recommend if you’re skeptical or not interested in nerdery1.

I used spaced repetition to make a professor think I was Talented at learning German. I used spaced repetition to ace tests on persnickety details of the C programming language and POSIX minutiae. I used spaced repetition to make people think I was naturally good at things when I wasn’t, and I used spaced repetition to improve my fiddly fundamentals in things I actually was naturally good at.

I am not going to provide any motivating cognitive science, hand-wavey or otherwise, because that argument has been made and I don’t need to.

Broadly, my experience led me to believe that spaced repetition is the best way to be able to remember things that you don’t actually care about in your gut. The things you care about you can always recall–a phrase in a foreign language that made you laugh hard, your crush’s class schedule, the core idea of your master’s thesis…

However, there’s a lot you may want to have memorized that you can’t emotionally care about until you need it: CPR cadences, interview answers, the word in a foreign language for “can opener”… When the moment strikes, you will care, but until then, how do you make it stick?

One thing people don’t often mention about SRS is that it is accommodating of human weakness. Too often the “construct from first principles based on Science” approach to reality believes it can paper over our failings and foibles, can ignore our socially determined differences, can build greenfield where structures exist. But when I go back to my Anki deck after weeks, months, years away–it sees that I am now forgetting what I once remembered, and it brings back what I need to relearn. After these gaps it’s interesting to see how I’ve changed. CS things I used to think I’d need to know now seem like laughable trivia; Latin phrases I’d wanted to memorize now have a different ring from the experiences I’ve had in between. Often I need to modify, delete, refocus the cards I’d made for myself. That’s okay. Your deck will change as you do.

In my university days, there was an email I sent around to everyone who asked about my setup2, so I’m going to reconstruct that resource here.

Here is a piece about why memorization in general is good. Its framing does not entirely jive with my own, but I’d consider it valuable for anyone questioning the whole premise.

what is it good for? (absolutely something)

  • vocabulary - in one’s native or acquired tongue, and especially when you need to memorize grammatical metadata along with a denoted meaning
  • memorizing poems/prayers/lyrics - you will need to split the poem up into little bits such that you are prompted with two lines and must recall the line that follows
  • dates, names, formulae - anything fiddly that you can’t rely on a holistic understanding for

tools

I have yet to find anything surpassing Anki.

The official docs have all the technical details you might need. When it comes to best practices for formulating cards, they link to this article by Anki’s competitor3.

why I am not sharing my deck

You really need to build your own. I know that seems like trite character-building nonsense, but the cards you make yourself are a thousand times more useful than decks you can import4. I advise being liberal about adding cards to your deck, but also very open to deleting them – it’s easy to realize that something is only worth memorizing with a certain cue, and quick to get rid of something you realize you didn’t care about at all.

An exception here might be decks made to follow along with a specific curriculum. I’m going back through Cambridge Latin and I am very pleased that someone already had a deck made with the vocabulary. Expect, however, that even these decks will require editing to adapt to your use if you want to start reversing cards5, as the cards aren’t always sensibly bidirectional.

tips

I recommend only segregating decks for language learning so you don’t have to note that this card with an English word on it is prompting you for the French translation rather than the Cantonese. General information should be all mixed up together. Similarly, if you’re not working with a premade deck built from simple to complex, I recommend adding new cards in random order.

However, one important thing I’ve only grokked after many years6 of this:

When you are using a premade deck, it is better to make sure you have no review backlog than to add new cards more aggressively. When you are making your own deck, it is better to ensure you are keeping up with new cards as you add them than to keep perfectly clear of review backlog.

Why is this?

Well, when you have a premade deck, the progress you can make is best measured as the value in the cards you manage to shift into the mature category. For this, you really need those reviews more than the new learning, and you should focus on giving the app all the feedback it can use about proper review intervals for each card.

When you’re making your own deck, however, the best thing you can do is handle those short review periods right up front while the value and context of the info on the card is clearest to you. This is partially for morale and partially so you don’t leave the utility of your deck in the yet-to-be-reviewed stack.

There are limits, of course. One can’t go on adding and adding new cards without reviewing. The guidance in the app settings that you should plan to review about ten times as much as you add isn’t bad. However, without mathing it out, I think it’s not bad to cram new cards as you add them on your days of active study, and to space out your active study as you’re able to knock down your reviews–without stressing about “did I get to all my 1 day interval cards within 1 day”7. When your daily reviews get quick again, that probably means it’s time to dive back into adding cards.

  1. Via Marijn, thanks Marijn!! ↩

  2. Typically after the first major test of the semester. This wasn’t to brag at the time, but bringing it up now definitely is. ↩

  3. The article isn’t necessarily worth reading all the way through except for number 4, I’d say. ↩

  4. Possibly in the same way that the act of paraphrasing in taking notes is valuable–figuring out how to best quiz yourself will give you the right start on memorization8. ↩

  5. For example, you can’t rely on languages to go perfectly back and forth. “Te amo” is best translated “I love you”, but “I love you” is best translated “te quiero”, mostly. I find it very frustrating when my cards tell me my correct answer wasn’t what they were looking for; YMMV. ↩

  6. Jesus, nearly a decade. I lost one of my Anki accounts in there somewhere, so I don’t have my old German deck, which would be a hoot to see again… Anyway, the point is I’m no greenhorn, you kids listen to me ↩

  7. Remember, if you delay reviewing a card past its scheduled time but still get it right, the algorithm will notice and boost that card out further into the future. ↩

  8. Phrased for those fond of other mnemonic methods: the design of the memory palace might be less initially important than the designing of it. ↩