A major limitation exists for spaced repetition software (e.g. Anki, SuperMemo, or other flashcard-style systems), in my experience using them for several years for long-term memory: it's neither necessary nor sufficient to use this software to learn certain topics.
Several excellent physics and math students I worked with have never used spaced repetition software, but were excellent at their studies because they consistently solved textbook problems. For them, they got the "repetitions" (aka exposure to facts and problems) via solving more new problems from the book nearly every day. Later problems in the textbooks would provide review of previous problems. This method can be far more effective than spending too much studying with spaced repetition software (which I have done in the past), as the time spent creating new cards and reviewing cards that are due comes at the expense of time spent solving new problems.
Ideally, you can perhaps find time for both activities. But in my personal experience, I learned mathematics more effectively by focusing primarily on textbook problems (reviewing older material through new problems) and spending a very limited amount of time with spaced repetition, versus even a fifty-fifty split between a textbook and spaced repetition that I've experimented with in the past.
In the past, I also spent too much time in the past remembering phrases and vocabulary when learning new languages, and not enough time practicing listening, writing, and especially conversation. Certain skills can only be efficiently developed by directly practicing them. While spaced repetition software remains an essential part to my language studies, it is very far from sufficient (even just a couple hours of conversation practice per week over three months, got me much further than primarily focusing on Anki/spaced repetition for six months).
Spaced repetition systems like Anki (though I moved to SuperMemo about a year back) are vital to my daily studies, but I've learned far more effectively by treating these systems as a supplement to more traditional and tested study methods that rely on active problem-solving. For technical fields, these are textbook problems from books by well-regarded authors, and for languages, these are conversation practice and writing articles that I request feedback on (in particular, teachers in small group classes have given me great, useful feedback on my writing).
Knowledge that is primarily conceptual (like almost all of math) generally does not benefit from spaced repetition. The learning involved is understanding -- a new concept may be hard to understand in the first place, but once you get it, you don't really forget it. Or you just need a super-quick refresher if you haven't touched it for a few months.
While knowledge that is primarily arbitrary-factual is the perfect candidate for spaced repetition -- mainly things like vocabulary, medical terms, and so forth. Just associating a largely arbitrary name for something. And indeed they are mostly useful for learning for exams. E.g. I used it to learn Chinese characters and could never have passed Chinese class otherwise. But on the other hand when I actually lived in another country that speaks a different language, spaced repetition isn't much of a help -- you learn vocab just by absorbing it day-to-day, like a kid does.
But the arbitrary-factual isn't they way people recommend using Anki. I really struggle to understand exactly what people are actually using it to learn. There are a lot of medical students and that makes sense, but that is... arbitrary-factual.
I've tried many times with other material, but I always get far more out of creating condensed notes and reviewing actual notes than quizzing random factioids in Anki.
I get that probably the idea is that creating prompts could guide me about which notes to review, but it seems easier to just review the notes. I think the people doing Anki are just putting the effort into making cards rather than into making notes. It would probably be nice if it were easy to convert notes into Anki but everything I've tried sucks and it just make taking each hour of taking notes into two hours of making notes and cards. I could have just reviewed notes for that hour. Closures are sort of okay but converting notes into closures is a bigger pain in the ass than it should be.
Maybe someone can make a LLM that takes some corpus of text and generates Anki cards of key points for review. Then I could just feed my notes in. But otherwise it really seems like a huge waste of time.
Quoting myself from four years ago: Dr Ali Abdaal runs a YouTube channel (about studying medicine at Cambridge University) and one of his videos is on "Evidence-based revision tips", with citations for the studies he's working from - here: https://www.youtube.com/watch?v=ukLnPbIffxE it's about 20 minutes.
He says that spaced repetition is effective, but basic repetition of re-reading, re-watching, re-listening is not effective.
Spaced repetition with "active recall" comes out significantly more effective - instead of exposing yourself to the same material over and over, challenge yourself to recall the material at the time when you're on the edge of forgetting it; the active mental effort of doing that appears to fix information in memory much more effectively than reading or hearing it again.
A consequence of that is his suggestion that notes and review material should not be facts you want to remember, but questions that will prompt you to think and recall what you want to remember. "Writing questions for yourself makes you engage in cognitive effort, and the more brainpower it takes to recall a fact, the better strengthened that connection seems to get, according to the evidence at least".
The thing I would counter is that "people studying medicine in medical school" is a bit of a bubble. This is particularly because the step of creating notes isn't necessary. There are many, many pre-built notes and study guides to select among. It's more about finding one that works for you and adapting it to the specific coursework.
The point being that if you are studying medicine the notes for review already exist and you just touch them up. You spend very little (if any) time actually creating them. I would certainly agree it's a waste of time to create your own notes in medicine.
I don't see how making and reviewing notes is that different from making and reviewing cards. The more important aspect of using a system like Anki is leveraging the Ebbinghaus forgetting curve. Maybe if you already have the habit of making, organizing and reviewing notes you don't need a system on top of it like Anki?
That's fine if you're a student and can consistently do textbook problems. What happens when you enter the workforce where you very likely will not use most of the mathematics and get married and have kids? You don't have time to solve those problems.
I had lost all hope till I started using spaced repetition to relearn what I'd forgotten. Is it as good as doing textbook problems? No. But it's better than not doing anything. With spaced repetition I can have gaps of months at a time and can still get back to the material without having to start all over.
N=1. I found that conceptual knowledge and eureka moments always result in substantial or at least non trivial changes in the way I parse and understand information, so much so that their integration is seamless, automatic and permanent.
Otoh, certain stuff that are not particularly important need frequent repetition, eg learning a foreign language that you are not actively using.
I am considering using it for leetcode problems and questions.
> Otoh, certain stuff that are not particularly important need frequent repetition, eg learning a foreign language that you are not actively using.
In that use case, you are much better off reading content in that language and watching content in that language. Vocabulary memorization as isolated activity makes sense only as additional activity if you have additional time on top of that.
I mainly use anki with cards that involve the vocab in some context. I found that the context enabled me to eventually understand without translating to any other language, I am still quite limited though. I think that falls into your suggestion of learning within some kind of broader context, which I agree.
It would definitely be hard to use Anki for conceptual learning that requires fresh problems to prevent memorizing answers. One could link to an outside site keyword categorizing problems matching specifics, I suppose..
Really though a lack of spaced repetition on practising conceptual problems is just as noticeable as the lack of spaced repetition on factual learning. Very few fields are so perfectly structured that you get practices on their earliest levels by doing later levels and then working in the field, i.e. plenty of mathematicians admit they no longer do arithmetic well.
I mean for math I think I would have benefited quite well from sticking in various known derivative forms into Anki. You can walk and chew bubble gum here!
I've been working on https://github.com/trane-project/trane for the past year or so, mostly to get around these limitations. I tried to find a way to use Anki or another existing software to aid my music practice, but I couldn't get it to work.
Some ways in which it's different:
- Dependencies are core to the system. For example, if I am learning a music piece, I want to start by learning small sections and only move on to larger sections when I am good enough at the small stuff, eventually ending with a final exercise that tests my performance of the whole piece. A lot of knowledge/skills follow that pattern, but I couldn't find a way to make Anki or SuperMemo understand this.
- It's meant for both memorizing stuff and practicing exercises. I have tested it with your exact example (math problems from textbooks). It works fairly well, but it's at a very early stage (you can look around at https://github.com/trane-project/trane-math, but it still needs a readme). So it's doing the same thing as the students you mentioned. The difference is that the scheduling is done automatically. Review of existing problems and addition of new ones happen without requiring planning or tracking from the student.
- There's an emphasis on generating the flashcards as text files, so they can be shared. I don't understand why people insist of remaking their own flascards every time. If someone wishes to learn guitar, for example, it's my hope they just download some courses and start learning without spending any time redoing flashcards. This design choice probably makes it harder to write the flascards, but it balances out once the flashcards are done and can be passed around.
A quick question that I could not answer from reading the Github page: How married is the learning process to those "courses" you describe? If I have a topic X and start with 0 knowledge about it, could I use trane to learn it if there is no premade course yet? I assume to create a course about it, I would have to have quite some knowledge about the topic already, to be able to think of exercises and rate them in terms of difficulty/hierarchy.
You are correct. Some domain knowledge is needed if you want to get the correct dependencies.
They can be changed without affecting the exercises so it's not set on stone. So you could try to make a lesson for the initial topics and try to adjust the dependencies as you go along.
But the best way is to get help from someone who knows them.
I’m switching my job next month. That means a lot of new acronyms to learn of software components, departments, and projects. Anki helps me learn them.
Sure, even for the task that it was designed for, learning Japanese, I mostly pursued a method of just writing stuff in a notebook over and over when I was getting that degree. But the specifics of Anki itself are less important an observation than the main one: despite the widespread conviction that rote learning is pointless it is actually extremely helpful to a wide number of fields of endeavor.
> Several excellent physics and math students I worked with have never used spaced repetition software, but were excellent at their studies because they consistently solved textbook problems.
Isn't it the case that only minority of students uses this software or flashcards in general anyway? I mean, of course it is possible to succeed without it, because overwhelming majority of students/learners are not using it.
The idea gets more interesting depending on your idea of success. If success just means passing the course, you definitely don't need spaced repetition (but then again, you can also maybe get by via cramming two or three days before each test and rushing assignments, though this wouldn't be a good experience).
If you define success as doing very well in the course (and remembering what you learn in the long-term), the idea gets more interesting because more people interested in these outcomes use flashcard systems. Anki is especially popular among medical students, so it's a tempting idea to consider that applying spaced repetition to subjects like mathematics or physics can also be useful.
To some people studying math or physics, the software can potentially be a nice aide, but in my individual experience, I understood and remembered concepts better by spending more time practicing problems and doing sample tests, versus spending more time using spaced repetition software.
That is the thing ... I don't recall best students doing flashcards all that much in cs, math and physics. I actually associated this more with groups of students who don't understand, so they memorize without understanding. When you understand the concept, it is also much easier to remember the thing.
That's a fair observation. My personal experience has a bias because I knew a math and computer science major who was hardworking and seemed bright at academics, and he was really into Anki.
Knowing him led to my impression that a decent number of math and physics students might have studied with spaced repetition without talking much about it, as he only mentioned it after he saw me studying with the software once. But in hindsight, I agree that it's more likely that math & physics students stick to the learning habits they developed earlier on—often through lots of practice tests and exercises.
Anki seems to be more of a natural progression for people in other fields like biology, where students might develop habits of using flashcards with software like Quizlet, where using software with better study schedules like Anki can be a natural progression.
My main motivation behind the switch was that I had too many daily reviews of very old cards that I knew well in Anki, after using the software very close to daily for several years. I read that SuperMemo handled older reviews better, and I was also optimistic at the time that the learning schedule really was better than Anki's (as SuperMemo uses a significantly later version of the algorithm that Anki is based on for scheduling reviews).
The software switch had its ups and downs. First, the downsides: a significant one-time cost included the time spent learning all the items from scratch, as the import of cards from Anki to SuperMemo didn't preserve the repetition history. Another one-time cost, though minor, was some friction setting up the software (it took an abnormal number of days to receive the activation code, which I eventually received after a follow-up; maybe the company had a problem with their system at the time).
Long-term downsides include the lack of easy image occlusion (aka, covering up parts of a labelled image and revealing just one label in separate flashcards). If I studied maps/diagrams with spaced repetition or anatomy like in my high school biology class, this would be a dealbreaker (though I suppose you could use keep using Anki along SuperMemo). In my experience, it's far easier to occlude images in Anki than in SuperMemo. Also, the Windows desktop version I use doesn't have a mobile version, which is a very significant downside. I'm now used to reviewing SuperMemo in the evening or when I think it's a good time, via software called Parallels to run a Windows virtual machine on a MacBook, but there's a lot more friction to starting a review session. To add to the friction, backups are a bit harder (though I've made it easier by setting up a custom keyboard shortcut to press the sequence of keys create a backup in a folder in cloud storage).
The main upside is that I do think (noting that I may have confirmation bias) that the flashcard scheduling really is effective and also more efficient. I no longer face significant numbers of very old reviews, and I do subjectively feel that I retain my cards better.
To add objectivity on effectiveness for accuracy, according to SuperMemo's data, I have a 97% retention rate on my French cards which contain a lot of very old cards, though I aggressively remake cards that become "leeches"; an 89% retention rate on my Spanish cards, which have much more newer cards; and a 93% retention rate on my mathematics/sciences/miscellaneous readings cards. Unfortunately, I didn't keep a records for Anki cards in comparison, and there may be other factors behind increased retention (if any) such as following better practices when making new cards [1]. I wish I had numbers for time efficiency, but I can confidently say that I don't dread spending time reviewing old cards (though once again, there may be some other factor).
On SuperMemo's other features: I also did try the "incremental reading" feature of SuperMemo, but ultimately, I borrowed some of the principles and stuck to a personal method of taking notes from different books, while switching between books (instead of only focusing on one subject a day) to help stay alert while studying the materials. There are also other features of SuperMemo for sleep tracking and task scheduling, but I didn't personally enjoy using them (I personally found that sleep tracking didn't account well for daylight savings, and I prefer other simple apps accessible by mobile devices for scheduling and task tracking).
To make a very long story short: I switched to try and reduce time spent during reviews, but ended up spending a lot more time setting up and getting used to the software. I didn't really mind the fiddling that much, as spaced repetition software was a sort-of hobby for me in the past, but for other people who'd just like to learn, Anki provides a far more direct way to try spaced repetition.
SuperMemo is effective for me now, as I don't spend much time at all fiddling with the software. But I'm not sure if I would enthusiastically recommend the switch to other people unless they're interested in spaced repetition software (and thus don't mind fiddling with it), and they also have some dissatisfaction with Anki in some way. In any case, with either software, I've found my studies to be more effective by treating spaced repetition as just a supplement to other forms of study that require active problem solving.
You're right that retrieval practice isn't optimal for all types of knowledge. But not all spaced repetition systems are flashcard systems.
Just under a month ago, I read about 'math academy' here on HN. One thing it does is sort of what it sounds like you envisage: surfacing exercises relating to concepts you might be about to forget.
I agree. I believe that it’s one of the things that sets apart great books. Lately I see this more often and think “Neat, they make me practice this old idea and make me build this new thing”. It makes an appreciate the work and thought that went into writhing the text much more.
I had the same experience. I tried Anki for 2 years to learn to crack programming interviews, and failed. Then I started doing leetcode problems regularly on the site and that helped me to get past my block.
Several excellent physics and math students I worked with have never used spaced repetition software, but were excellent at their studies because they consistently solved textbook problems. For them, they got the "repetitions" (aka exposure to facts and problems) via solving more new problems from the book nearly every day. Later problems in the textbooks would provide review of previous problems. This method can be far more effective than spending too much studying with spaced repetition software (which I have done in the past), as the time spent creating new cards and reviewing cards that are due comes at the expense of time spent solving new problems.
Ideally, you can perhaps find time for both activities. But in my personal experience, I learned mathematics more effectively by focusing primarily on textbook problems (reviewing older material through new problems) and spending a very limited amount of time with spaced repetition, versus even a fifty-fifty split between a textbook and spaced repetition that I've experimented with in the past.
In the past, I also spent too much time in the past remembering phrases and vocabulary when learning new languages, and not enough time practicing listening, writing, and especially conversation. Certain skills can only be efficiently developed by directly practicing them. While spaced repetition software remains an essential part to my language studies, it is very far from sufficient (even just a couple hours of conversation practice per week over three months, got me much further than primarily focusing on Anki/spaced repetition for six months).
Spaced repetition systems like Anki (though I moved to SuperMemo about a year back) are vital to my daily studies, but I've learned far more effectively by treating these systems as a supplement to more traditional and tested study methods that rely on active problem-solving. For technical fields, these are textbook problems from books by well-regarded authors, and for languages, these are conversation practice and writing articles that I request feedback on (in particular, teachers in small group classes have given me great, useful feedback on my writing).