Anki and Spaced Repetition

Notes on working with a new system, 5½ months along.

Assumed Audience: Generally, people interested in memory, learning, and tools for the same; but specifically, people who already know how spaced repetition systems work (I’m not going to explain it).

Epistemic Status: A personal experience report. Your mileage may vary.

Does it work?

I started working with Anki almost exactly 5½ months ago (on May 15). I can safely say, this far along, that it is easily the most effective tool I have ever used for deeply committing things to memory. In that span, I have relearned well over a thousand Greek and Hebrew vocabulary words — and, frankly, learned them better than I had learned them the first time. I have gotten to the point where I can recognize all the white key” pitches in the middle octave of the piano nine times out of ten. I have internalized a bunch of logic and set theory symbols, so that I now recognize them when reading papers. I learned what a bunch of terms from music production mean.

In short, spaced repetition works.

But ugh

Experientially, committing to using Anki (or any tool like it, I expect!) kind of stinks. It requires daily review, especially if working on relatively large numbers of new cards. And by relatively large” I mean more than 3 or 4 a day.” Review time piles up very quickly if you add too many cards too quickly. I started out trying to learn 20 cards a day. Within a week or so I dropped that back to 10. Within a month and change, I had it down to 4 – 5 cards a day, just to keep it sustainable.

Fundamentally, spaced repetition only works if you actually do the repetition part of that, and in general that means you have to make time for it… daily. That’s not necessarily a big deal if the time you spend on it is 5 or 10 minutes tops. When it’s climbing up to 30, that becomes a very serious investment. Many people who write about the wonders of spaced repetition write about it in the context of their day-to-day work, and I can see how that would make it far more doable. For me, that means taking an extra 30 minutes out of my day beyond work. That’s time I can’t spend on other things — something has to give, and if I refuse to cut into time with my wife and daughters, that means that reading, composing music, or writing all have to get cut instead.

I found a balance here eventually, but it took a bit. If you pick up a spaced repetition system, start slow and build up to a level that feels consistently manageable. That means it has to be manageable even on days when you’re exhausted. It also has to be something you choose to build into your schedule. I continue to struggle with it, to be honest, because I haven’t found a good home” for it in my day-to-day rhythms.

I also found that different kinds of memory require different levels of effort. Trying to memorize three verses of Scripture is very different than trying to memorize three ideas from music production. That’s because learning an idea is not the same as memorizing a text. On the one hand, this should probably be obvious. On the other hand, I just had to scrap most of what I had planned for last week and this week and start over on it because I got overwhelmed by it and so stopped doing it and so lost a week of reviewing of everything else and then had to catch up on 160 cards tonight, and that took me 25 minutes all on its own. I will be resetting this bit of Scripture memory work tomorrow, but… I will be doing one verse a day instead of three.

Tweaking

A few weeks in, I found and took some very good advice about working with Anki specifically. (Other spaced repetition systems may work differently.) Out of the box, if you miss a card, Anki resets the time to your next review of that card back to the lowest repetition value: one day. This is kind of weird, and in my opinion basically wrong. The whole point of spaced repetition is that you come back to a given card when it should be right at the edge of your ability to remember it, and the work of remembering it then solidifies it much more robustly. As noted above: this is true! It works!

But… if that’s the case, then it doesn’t make any sense at all for a card you can’t quite remember to go back to being reviewed just a day later. If you need one prompt to remember it, and then it’s solid again, it should go back to longer-term repetition, rather than starting from scratch. Starting over is wasted time and effort! This wasn’t just hypothetical for me. I found it super annoying to review a card the next day that was too easy, having reviewed it just the day before. So I changed Anki’s settings to halve the next schedule repetition on a miss, rather than resetting it entirely.

This means that if I review a card, miss it, and then remember it when reviewing again later in the session, I’ll see it again on some reasonable time out — if it would have been a month, it’ll be two weeks, etc. By contrast, with a card that actually is giving me problems I can end up back in a spot where I will end up resetting it all the way if I fail over and over again in a given session.

Anki also has a system for cards it calls leeches”: the idea is that if you fail on a given card over and over after it graduates out of the new” category, then the card is costing too much energy and isn’t worth your time. Maybe that is true for some things for some people. For me, if I have put it in my list of cards to learn, I am judging it worth my time to learn, and the fact that it is hard or annoying is quite irrelevant. I can choose to pull a card out if I want to; but the app should never do that for me. So I turned the leeches” functionality off.

So…

I’m going to keep using Anki, because, again: it works. It has been really nice to have solidified so much Greek and Hebrew vocabulary and grammar again — and indeed, more than theoretically nice: actually actively helpful in things like prepping for preaching a guest sermon at a friend’s church. I’m using it in a more measured way than when I started, though, and I’m still trying to find just the right slot to fit it into my schedule.