Consider this. You make five flashcards for Russian vocabulary words. Russian word on one side, English on the other. Review those cards 20 times a day for a week. At the end of the week, you'll know those words really well. Now deposit those cards in the bank with instructions to keep them hidden from you for a whole year. Come back one year later and how well you will remember those words? You probably won't remember them at all. You reviewed each card 140 times, for a total of 700 flashcard reviews, and all for nothing. Pretty discouraging, isn't it? But what if I told you that you could review each card just 20 times total, and remember them all perfectly at the one year mark? Sound too good to be true? Welcome to the wonderful world of Spaced Repetition Systems. It's a world psychologists have known about for over a hundred years, but the calculations involved were too difficult to be practical until recently. Science fiction writers have long fantasized about a future where we can quickly learn vast amounts of knowledge using technology. We stand now upon the brink of that future, and that technology is the Spaced Repetition System.
A Spaced Repetition System, or SRS, is a program that lets you review flashcards. But here's the catch. Each time you review a flashcard, you rate yourself on it, based on how easy or difficult the card was. The program uses all kinds of advanced calculations and sophisticated algorithms to determine the optimal possible way to show you those cards. In the Russian vocabulary example above, reviewing each card 20 times a day for a week is very far from optimal. It would be much more efficient to review them a few times to learn them, and then wait a while, and then review them once more, and then wait even longer, review them again, and so on. The underlying discovery from lab psychology is that, if the timing is done right, each time we review a card we can go longer before we need to review it again. The difficulty is calculating that timing just right; that's where the SRS flexes its muscles. When the spacing is done right, it takes very few reviews before you can go months or even years between reviews for that card.
So what's the catch? Surely these Spaced Repetition Systems cost thousands of dollars or give you memory cancer? Nope, some of the best SRS's are free, and I've been using them for years without developing any memory cancer (or at least, none that I can remember!) The most popular free SRS is Mnemosyne, and another one of my personal favorites is Anki. The latter is relatively new and being developed extremely actively. The only possible drawback to an SRS is that it's meant to be reviewed every day. You can take days off here and there, but in general, you need to review your cards daily, in order for the SRS's calculations to work their magic. I personally consider this almost a good thing, because it has really given me a boost to my self-discipline and it's good to have a reliable anchor like this in my otherwise hectic life. Besides, with things like languages, you should be studying every day anyway.
Note: Thanks to Tobberoth from this excellent forum thread for the awesome mental experiment about the Russian flashcards, which I modified slightly.
HOW MUCH DAILY WORK IS SPACED REPETITION?
The amount of work you put into spaced repetition depends how much you want to get back out of it. The thing is, as you review any particular card, it'll become more and more spaced, needing to be reviewed less and less often. Right now my Japanese language deck has over 9,000 cards, but most of those cards aren't scheduled for review for over 100 days.
A general rule of thumb is that on any particular day, the overwhelming majority of the cards scheduled for review in an SRS will be the cards you added most recently. Therefore, the amount of daily work is approximately equal to how many cards you've recently added. In particular, if you just need to memorize some finite amount of information with a definite end, like the capitals of all the nations of the world, it'll be a lot of work for a short while and then rapidly dwindle down until it's almost nothing.
If you want to put in the work, this means that an SRS will let you learn at a staggering rate. It's almost illegal. In fact, I'm pretty sure if the government ever finds out just how efficiently and optimally people are learning with SRS's, they'll ban the things for being too good for the common citizens :)
MAINTAINING WHAT YOU'VE LEARNED
As great as they are for learning things, the real power of the SRS is in maintaining what you've learned. Take languages, for example (one of the most commonly studied things on an SRS). What stops most people from learning lots of languages isn't that each individual language is hard. Languages aren't that hard, after all, babies learn them all the time and babies suck. No, what really stops people from becoming polyglots is that it'd take too long to maintain all those languages. You could learn German, then learn French, but by the time you've learned French, you've forgotten three quarters of your German! Or keep studying German, but then it takes twice as long to learn French, and four times as long to learn the next language after that...
With spaced repetition, maintaining what you've learned is easy. And it gets easier and easier the longer you've been at it. I could stop studying any new Japanese today, swear it off, cold turkey, and never listen to or read a word of real Japanese again in my life, but if I keep doing my daily reviews, my Japanese knowledge will hover right around where it's at today. And within one or two months after swearing it off, my daily flashcard review will be down to maybe 20 cards a day. Within one year, it'd be in the single digits.
OTHER ADVANTAGES OVER PAPER FLASHCARDS
Besides the huge advantage of optimized learning, there are other advantages to doing your flashcards on the computer. These advantages would hold even if the program just naively showed you the flashcards in order, without any sophisticated scheduling computations.
* Handle extremely huge decks without filling your closet with shoeboxes. When I first found out about Spaced Repetition Systems, my office had a giant brick of paper Japanese flashcards, and that was when my Japanese language study was still extremely young (in other words, I probably had less than 1,000 actual cards then). And those cards cost money, too! If it's a dollar for a hundred blank flashcards, my Japanese deck right now would cost about $90 just for the raw paper, not counting ink or storage. When my dad was in grad school, he actually made his own flashcards, cutting them out of scratch paper to save money, and of course that added up to many hours just cutting paper. And good luck managing such a huge deck! That would be a nightmare!
* Backup and transfer. With flashcards on a Spaced Repetition System, you can back the cards up, even back them up online so you keep your deck even if your whole house burns to the ground. You can also email your deck between work and home and school and not have to carry it around everywhere (no more bulk bags of rubber bands!) Anki even has a "synchronization" feature that makes this all automatic, if you want.
* Detailed statistics. With a few clicks, you can find out all kinds of cool statistics about your cards. Some SRS's (like Anki) will even create graphs and pie charts and stuff like that.
* Multimedia cards. You can put pictures and even audio files on flashcards. Audio files are ridiculously useful for language flashcards, it's like God's gift to flashcards. Some SRS's even let you "drag and drop" pictures right into the program, making it quick and easy to make the things.
* Community decks. As SRS's become more popular, people are starting to publish pre-made decks, free for you download. Then you only need to review the flashcards, you no longer have to even make them! One SRS site which specializes in community cards is Flashcard DB, which has thousands of community-made decks for free.
COMMON MISCONCEPTIONS ABOUT SPACED REPETITION SYSTEMS
Misconception #1: There are only certain fixed things that you can download cards for.
The truth: This is true of some specific spaced repetition systems which are built for certain specific tasks, like the SRS at Reviewing The Kanji, which only lets you study the Japanese characters. But all the big SRS's, like Mnemosyne and Anki, allow you to make your own cards for anything you want. Basically, you have as much freedom as you do with paper flashcards, in fact you have even more freedom since you get all the advantages I talked about above.
Misconception #2: "I accidentally saw Vocabulary Word XXX outside my SRS studies, oh no, now the spacing for that word will be all messed up!"
The truth: The point of the spacing algorithms is to show you the card as rarely as possible and still keep it in your brain. Of course if you studied the card even more, you'd remember it even better. Otherwise, you and me would have totally forgotten the Roman alphabet by now, I mean we see those letters all the time and never pay attention to the spacing between reviews! The point about the SRS, though, is that you don't have to see the material between reviews in order for it to stick.
Misconception #3: Spaced Repetition Systems are inferior to paper flashcards because you can't take the cards with you!
The truth: It's true you can't lug a desktop computer around everywhere, but many SRS programs are now available on various PDA's and whatever other handheld devices are out there. In any case, many SRS's let you do your review over the internet, so any handheld access to the internet will work for SRS'ing too.
SPACED REPETITION FOR BUSY PEOPLE
When I took a month long break from SRS'ing, I came back to a big pile o' scheduled cards. I was too busy to plow through them all, but at the time I only wanted to maintain the knowledge, I wasn't actively adding new cards at the time. I discovered the perfect compromise, which is to only review N flash cards per day, regardless of how many were scheduled. For me, N was 100, and it worked pretty well. I can review 100 flashcards in half an hour or less pretty easily.
So what's the lowest N could theoretically go? In principle, any positive number will do, no matter how small. It might take a very long time to get through the whole deck, longer and longer the smaller N is, but any positive N will work. The reason is that, even if you're initially moving backward (doing fewer reviews than the program is adding every day), you will be accelerating forward because of spacing. You might be making negative net progress, but as you make what progress you do, you'll be increasing the intervals for those cards, so that eventually fewer and fewer new reviews will be scheduled, and you can eventually start making the ground back up.
The same idea would work if you devoted N minutes a day of study to reviewing cards, instead of reviewing N cards per day.
Even if you have only ten free minutes every day, Spaced Repetition can still work for you!
USING SPACED REPETITION SYSTEMS FOR SCHEDULING
If you're creative, you can use Spaced Repetition for more than just memorizing facts. I've been experimenting with using Mnemosyne to diversify my music-listening.
Basically, if you listen to the same song over and over, it gets tiresome. And if you haven't listened to a song in awhile, it sounds better. Hmmm, this sounds analogous to something, doesn't it...
I'm experimenting with putting song names in Mnemosyne, and then listening to whatever song is "up for review", and then rating it based on how good it sounds. A better-sounding song, I'll rate worse, meaning there'll be less space before it's scheduled again. A really good-sounding song, I'll outright fail, meaning its space goes back down to 0.
In the long run, spacing will make fewer and fewer things come up for review, forcing me to discover new music and broaden my music reality. This is still a pretty new project, but I might write a full-fledged article about it here later, so keep in touch!
MORE ON THE SUBJECT OF LEARNING AND MEMORY
Sentence Mining is the world's most advanced technique for second language acquisition. It makes use of the Spaced Repetition System, and basically lets you learn a language with the natural exposure of a child but the efficiency of an adult.
The French Revolution: French in 30 Days. In order to learn more about language and how it's learned, and about memory in general, I'm doing a 30-day French language-learning challenge. On day 1, I knew absolutely nothing about French. Each day has a daily writeup. I use Spaced Repetition and Sentence Mining extensively and write about them in great detail.
One of the most difficult things you can memorize is the set of Japanese characters. In my book review of James Heisig's "Remembering The Kanji", I write about the most cutting edge techniques for making this feat doable. And yes, Spaced Repetition Software plays a big part.
When you're a self-teacher, life suddenly becomes a lot simpler. In my article, Autodidact: Be A Self-Teacher, I talk about being a self-teacher and some ways you can train this meta-skill. Being good with Spaced Repetition Systems will make self-teaching a lot easier. Plus, you kinda have to be a self-teacher because SRS's are so cutting edge and elite, they haven't made their way into the classroom yet.
A Spaced Repetition System, or SRS, is a program that lets you review flashcards. But here's the catch. Each time you review a flashcard, you rate yourself on it, based on how easy or difficult the card was. The program uses all kinds of advanced calculations and sophisticated algorithms to determine the optimal possible way to show you those cards. In the Russian vocabulary example above, reviewing each card 20 times a day for a week is very far from optimal. It would be much more efficient to review them a few times to learn them, and then wait a while, and then review them once more, and then wait even longer, review them again, and so on. The underlying discovery from lab psychology is that, if the timing is done right, each time we review a card we can go longer before we need to review it again. The difficulty is calculating that timing just right; that's where the SRS flexes its muscles. When the spacing is done right, it takes very few reviews before you can go months or even years between reviews for that card.
So what's the catch? Surely these Spaced Repetition Systems cost thousands of dollars or give you memory cancer? Nope, some of the best SRS's are free, and I've been using them for years without developing any memory cancer (or at least, none that I can remember!) The most popular free SRS is Mnemosyne, and another one of my personal favorites is Anki. The latter is relatively new and being developed extremely actively. The only possible drawback to an SRS is that it's meant to be reviewed every day. You can take days off here and there, but in general, you need to review your cards daily, in order for the SRS's calculations to work their magic. I personally consider this almost a good thing, because it has really given me a boost to my self-discipline and it's good to have a reliable anchor like this in my otherwise hectic life. Besides, with things like languages, you should be studying every day anyway.
Note: Thanks to Tobberoth from this excellent forum thread for the awesome mental experiment about the Russian flashcards, which I modified slightly.
HOW MUCH DAILY WORK IS SPACED REPETITION?
The amount of work you put into spaced repetition depends how much you want to get back out of it. The thing is, as you review any particular card, it'll become more and more spaced, needing to be reviewed less and less often. Right now my Japanese language deck has over 9,000 cards, but most of those cards aren't scheduled for review for over 100 days.
A general rule of thumb is that on any particular day, the overwhelming majority of the cards scheduled for review in an SRS will be the cards you added most recently. Therefore, the amount of daily work is approximately equal to how many cards you've recently added. In particular, if you just need to memorize some finite amount of information with a definite end, like the capitals of all the nations of the world, it'll be a lot of work for a short while and then rapidly dwindle down until it's almost nothing.
If you want to put in the work, this means that an SRS will let you learn at a staggering rate. It's almost illegal. In fact, I'm pretty sure if the government ever finds out just how efficiently and optimally people are learning with SRS's, they'll ban the things for being too good for the common citizens :)
MAINTAINING WHAT YOU'VE LEARNED
As great as they are for learning things, the real power of the SRS is in maintaining what you've learned. Take languages, for example (one of the most commonly studied things on an SRS). What stops most people from learning lots of languages isn't that each individual language is hard. Languages aren't that hard, after all, babies learn them all the time and babies suck. No, what really stops people from becoming polyglots is that it'd take too long to maintain all those languages. You could learn German, then learn French, but by the time you've learned French, you've forgotten three quarters of your German! Or keep studying German, but then it takes twice as long to learn French, and four times as long to learn the next language after that...
With spaced repetition, maintaining what you've learned is easy. And it gets easier and easier the longer you've been at it. I could stop studying any new Japanese today, swear it off, cold turkey, and never listen to or read a word of real Japanese again in my life, but if I keep doing my daily reviews, my Japanese knowledge will hover right around where it's at today. And within one or two months after swearing it off, my daily flashcard review will be down to maybe 20 cards a day. Within one year, it'd be in the single digits.
OTHER ADVANTAGES OVER PAPER FLASHCARDS
Besides the huge advantage of optimized learning, there are other advantages to doing your flashcards on the computer. These advantages would hold even if the program just naively showed you the flashcards in order, without any sophisticated scheduling computations.
* Handle extremely huge decks without filling your closet with shoeboxes. When I first found out about Spaced Repetition Systems, my office had a giant brick of paper Japanese flashcards, and that was when my Japanese language study was still extremely young (in other words, I probably had less than 1,000 actual cards then). And those cards cost money, too! If it's a dollar for a hundred blank flashcards, my Japanese deck right now would cost about $90 just for the raw paper, not counting ink or storage. When my dad was in grad school, he actually made his own flashcards, cutting them out of scratch paper to save money, and of course that added up to many hours just cutting paper. And good luck managing such a huge deck! That would be a nightmare!
* Backup and transfer. With flashcards on a Spaced Repetition System, you can back the cards up, even back them up online so you keep your deck even if your whole house burns to the ground. You can also email your deck between work and home and school and not have to carry it around everywhere (no more bulk bags of rubber bands!) Anki even has a "synchronization" feature that makes this all automatic, if you want.
* Detailed statistics. With a few clicks, you can find out all kinds of cool statistics about your cards. Some SRS's (like Anki) will even create graphs and pie charts and stuff like that.
* Multimedia cards. You can put pictures and even audio files on flashcards. Audio files are ridiculously useful for language flashcards, it's like God's gift to flashcards. Some SRS's even let you "drag and drop" pictures right into the program, making it quick and easy to make the things.
* Community decks. As SRS's become more popular, people are starting to publish pre-made decks, free for you download. Then you only need to review the flashcards, you no longer have to even make them! One SRS site which specializes in community cards is Flashcard DB, which has thousands of community-made decks for free.
COMMON MISCONCEPTIONS ABOUT SPACED REPETITION SYSTEMS
Misconception #1: There are only certain fixed things that you can download cards for.
The truth: This is true of some specific spaced repetition systems which are built for certain specific tasks, like the SRS at Reviewing The Kanji, which only lets you study the Japanese characters. But all the big SRS's, like Mnemosyne and Anki, allow you to make your own cards for anything you want. Basically, you have as much freedom as you do with paper flashcards, in fact you have even more freedom since you get all the advantages I talked about above.
Misconception #2: "I accidentally saw Vocabulary Word XXX outside my SRS studies, oh no, now the spacing for that word will be all messed up!"
The truth: The point of the spacing algorithms is to show you the card as rarely as possible and still keep it in your brain. Of course if you studied the card even more, you'd remember it even better. Otherwise, you and me would have totally forgotten the Roman alphabet by now, I mean we see those letters all the time and never pay attention to the spacing between reviews! The point about the SRS, though, is that you don't have to see the material between reviews in order for it to stick.
Misconception #3: Spaced Repetition Systems are inferior to paper flashcards because you can't take the cards with you!
The truth: It's true you can't lug a desktop computer around everywhere, but many SRS programs are now available on various PDA's and whatever other handheld devices are out there. In any case, many SRS's let you do your review over the internet, so any handheld access to the internet will work for SRS'ing too.
SPACED REPETITION FOR BUSY PEOPLE
When I took a month long break from SRS'ing, I came back to a big pile o' scheduled cards. I was too busy to plow through them all, but at the time I only wanted to maintain the knowledge, I wasn't actively adding new cards at the time. I discovered the perfect compromise, which is to only review N flash cards per day, regardless of how many were scheduled. For me, N was 100, and it worked pretty well. I can review 100 flashcards in half an hour or less pretty easily.
So what's the lowest N could theoretically go? In principle, any positive number will do, no matter how small. It might take a very long time to get through the whole deck, longer and longer the smaller N is, but any positive N will work. The reason is that, even if you're initially moving backward (doing fewer reviews than the program is adding every day), you will be accelerating forward because of spacing. You might be making negative net progress, but as you make what progress you do, you'll be increasing the intervals for those cards, so that eventually fewer and fewer new reviews will be scheduled, and you can eventually start making the ground back up.
The same idea would work if you devoted N minutes a day of study to reviewing cards, instead of reviewing N cards per day.
Even if you have only ten free minutes every day, Spaced Repetition can still work for you!
USING SPACED REPETITION SYSTEMS FOR SCHEDULING
If you're creative, you can use Spaced Repetition for more than just memorizing facts. I've been experimenting with using Mnemosyne to diversify my music-listening.
Basically, if you listen to the same song over and over, it gets tiresome. And if you haven't listened to a song in awhile, it sounds better. Hmmm, this sounds analogous to something, doesn't it...
I'm experimenting with putting song names in Mnemosyne, and then listening to whatever song is "up for review", and then rating it based on how good it sounds. A better-sounding song, I'll rate worse, meaning there'll be less space before it's scheduled again. A really good-sounding song, I'll outright fail, meaning its space goes back down to 0.
In the long run, spacing will make fewer and fewer things come up for review, forcing me to discover new music and broaden my music reality. This is still a pretty new project, but I might write a full-fledged article about it here later, so keep in touch!
MORE ON THE SUBJECT OF LEARNING AND MEMORY
Sentence Mining is the world's most advanced technique for second language acquisition. It makes use of the Spaced Repetition System, and basically lets you learn a language with the natural exposure of a child but the efficiency of an adult.
The French Revolution: French in 30 Days. In order to learn more about language and how it's learned, and about memory in general, I'm doing a 30-day French language-learning challenge. On day 1, I knew absolutely nothing about French. Each day has a daily writeup. I use Spaced Repetition and Sentence Mining extensively and write about them in great detail.
One of the most difficult things you can memorize is the set of Japanese characters. In my book review of James Heisig's "Remembering The Kanji", I write about the most cutting edge techniques for making this feat doable. And yes, Spaced Repetition Software plays a big part.
When you're a self-teacher, life suddenly becomes a lot simpler. In my article, Autodidact: Be A Self-Teacher, I talk about being a self-teacher and some ways you can train this meta-skill. Being good with Spaced Repetition Systems will make self-teaching a lot easier. Plus, you kinda have to be a self-teacher because SRS's are so cutting edge and elite, they haven't made their way into the classroom yet.
6 comments:
So, if you've got 9,000, how fluent would you say you are in Japanese? What can you do and what can you not do right now?
Just curious...
Wow--thank you for posting about this! I read an article months ago about a man who discovered/invented such a system and was fascinated, but the article implied that software for personal use was a long way off. I'm so pleased to hear that it's available now. I can't wait to use it to help learn Welsh.
P.S. I just started a new blog for 30-day trials, inspired in part by your writings on self-discipline--I'd love it if you would check it out and leave me a comment! http://30daysbetter.blogspot.com
Jakks: I'm pretty good at Japanese. I can do basic conversation, I can follow romance-comedy anime without subtitles, I can make some sense out of most written Japanese. Kanji-wise, I can write 1775 kanji. I'm kind of at a point right now where I see further study as a luxury purely for my own fun: I've already gone to Japan and survived 30 days, even got inducted into a crazy buddhist cult for fun. So unless I go allll the way to perfect fluency, it's just the difference between "neat gaijin who can speak some japanese" and "neat gaijin who can speak some (more) japanese". I'm actually a little bit sad to have this realization, it's almost like I'm parting with an old friend.
Over 9000! I'm around 6,500 in Japanese and 1001 in French. I'm not that good though, only been at it a year. I think I need another year of listening and speaking to put me over the top.
Hello Glowing face man,
Your article (i consider it already a "real" article) is very interesting and i am thinking about how to use spaced repetition for University learning.
Is it possible to use the method, without actually having to type in all the 'to-learn' knowledge? Let's imagine an exmple:
I have read a book with 10 chapters, each including 5 sub-topics, from which i want to remember the main points. Is it possible to make the program remind me when i have to review the main points of a specific sub-topic and to give a grade on it?
Or are there too many facts, that cannot be given a common grade. For example if i remember only 3 of 5 main points, than the 2 missing points would need further repetition, then the others. Would this kind of disturb the whole method, or do you think it would still work?
greetings, Kalki
Kalki: you could make the "question" side of the card be the name of a subsection, and leave the "answer" side blank. Then when the card comes up for review, review the subsection and rate it like: 0="I don't understand the section at all", 3="This seemed like a good time to review", 5="I think this scheduled review was too soon and the next review should take a lot longer". I haven't done that myself, it would be an interesting experiment!
Post a Comment