Note to writers: unless you're writing only for an audience that is already familiar with your terminology, it's helpful to spell out acronyms and initialisms the first time you use them.
I skimmed this entire article out of curiosity. It uses the term SRS twenty times but never says what it means. I went back to the top and noticed the "not a prerequisite" link, so I looked at that article too. About six paragraphs in, the fourth mention of SRS finally gave me the key: "Spaced Repetition Software (SRS)".
This tiny bit of knowledge gave me the context to understand what the articles were talking about.
If you put definitions like this at the top of every article, you will have more informed readers who can better appreciate your insights.
It's worse than that, because even for someone who knows about SRS (via Anki), this is circuitous and hard to read.
Anyone who's a fan of incremental reading/super memo would do us a favor by writing a new introduction that's more straightforward and credible. The fact that the introductions to incremental reading have this feel of "read these hundred things to grasp the genius of incremental reading" is a huge problem.
Exactly. Not only did I not know the terminology, in this case I didn’t know either product/software/service either.
I had to find the “in case you missed my earlier post on” then in that the “my initial review” post then still had to figure out what he was talking about. OH.... it’s learning by repetition, updated flash cards, got it in three words.
This was written by an engineer for sure! A short statement and purpose for people who don’t magically have all the information the writer has would have gone a long way.
I’m OP. Thank you for the feedback! I have added a quick terminology guide at the top of each article.
> A short statement and purpose...
I will keep this in mind when writing articles in the future.
The fundamental flaw in the author's argument for using SuperMemo is that there is a ton of value in manually creating the decks yourself. When you write out an item in pencil on paper, then copy it to Anki and create a card, you're creating a memory and physical association with the creation of that card. When you just cloze pre-made selections of reading material, you're skipping this entire process.
This is the same reason that downloading pre-made decks from Anki's website doesn't really work.
So true. There just aren't any shortcuts when it comes to learning a new language, and while creating cards/decks in Anki is a huge chore at times, it's also super effective in jack-hammering that information into your brain for long-lasting recallability.
Yes, I think decks have to be personalized. But I don't agree the argument that "writing out in pencil then transferring back to a Spaced Repetition Software" is better than "pre-made selections of reading material". For Incremental Reading, you have to select the important parts; there's no pre-made selection. It's just like highlighting; the material isn't pre-highlighted.
I think the most important thing is the thought process: thinking which part is important and which isn't. The medium, whether you are doing it in paper and pencil is no difference than doing it digitally.
>"Studies have shown that taking notes by longhand will help you remember important meeting points better than tapping notes out on your laptop or smartphone."
By "doing it" I meant the thought process, the selecting process of what's important, not the medium in which you do your "whole studying".
Also that study means that taking notes by hands is better than with a computer. It doesn't mean learning with paper and pencil is better than with a computer, especially when we involve other things like a spaced repetition software.
I'm really skeptical of these sorts of memorization systems, not because I think they don't work (I think they do), but because I see it as a brain "hack" that goes against the grain.
What do I mean by this? I think of our memories as a kind of most-frequently-used (MFU) cache. We remember best when we use the information all the time. So when we use this sort of spaced repetition we're trying to artificially "warm up" the cache in our brains. We're not trusting our brains to remember the things that are most important in our lives.
If you're trying to learn a language, you should learn the way children learn their first language: by speaking. Talk to people in that language constantly. If you don't know anyone who speaks the language, what are you learning it for? If you really need to learn the language (for work?) then find somebody who speaks it and arrange to meet them for coffee on a regular basis.
If you're trying to memorize something else, such as theorems in math or standard library functions in a programming language, I'd advise against it. Learn the theorems by applying them and trying to prove stuff. Learn the library functions by doing actual programming.
In the long run, if your knowledge is situated within a context and connected to other things you know (because you learned by doing) then I believe you'll understand it much better. Memorizing by rote, out of context, will help you to recall facts which often don't have any meaning to you.
> What do I mean by this? I think of our memories as a kind of most-frequently-used (MFU) cache. ... We're not trusting our brains to remember the things that are most important in our lives.
Absolutely not. I would never trust my brain to remember the things that are most important in my life. When is my wife's birthday? That happens once a year. My neurobiology is simply unsuited to triggering something at a specific time once a year. Plus our memory is not an MFU cache, not even close. Abandon all computer science metaphors and engage with the actual biology.
> If you're trying to learn a language, you should learn the way children learn their first language: by speaking.
That's like saying that athletes shouldn't do strength training or drills for specific skills. Mind you, I have learned a number of languages for real world use, and SuperMemo is one of my core tools for doing so.
> If you're trying to learn a language, you should learn the way children learn their first language: by speaking. Talk to people in that language constantly.
I agree that this is ideal, but it's also a huge synchronous time sink for the parties involved. When you're learning with flash cards, you can do it on your own time, in a more directed manner with trackable, linear improvement. In less structured environments, you don't get the breadth of vocabulary. In the beginning, you also end up asking "what does _ mean?" over and over again, or scrambling to write it down and look it up later, all while failing to grok the conversation happening around it.
Vocabulary alone is a huge barrier to entry in a language and, in my opinion, worth attempting to memorize in order to tackle a language.
Have you ever learned a language in your 20s? If so, how good are you? How long did it take you?
I personally have tried to learn many languages in my 20s. I’ve became good at figuring out what are the shortcuts to kickstart learning. The best one I found is to learn as many words as possible + simple grammar + useful verbs. If you have the time learn hundreds per day with memory tricks. Once you know enough start trying to have conversations. Grammar is not really important in the beginning but you need to learn conjugations (if there are any) at some point early.
This is a much better approach vs trying to have convos when you know peanuts. If you’re planning to spend some weeks in a foreign country this is what I would advise to you. You are going to learn very little if you get there without having done that
> If you're trying to learn a language, you should learn the way children learn their first language: by speaking.
I respectfully disagree here. After using the methods/strategies in the Fluent Forever book/videos/etc... (of which SRS is but one), while this strategy will work, its ultimately a longer road than necessary.
Kids don't get the chance to learn any other way than by observation and learning what not to do by inference. They're also very willing to make mistakes to tune their understanding. Adults tend not to be, taking the kids learning strategy as an adult is both ignoring the fact that you can take advantage of being an adult that can take in information differently, AND can speak to natives in a more structured way to much more easily learn grammar.
You'll still need to practice with native speakers no doubt, but knowing/learning words based on their frequency in the target language is something that you as an adult can take advantage of that kids realistically have to infer.
Immersion is great, but isn't ultimately a realistic option for most adults.
> but because I see it as a brain "hack" that goes against the grain.
The current, modern, and typical grain if you will: language learning is vocabulary, vocabulary and maybe conversation (maybe).
Although early on in childhood, as we develop "receptive language", we are going against this grain naturally by learning phonetics first then moving onwards to vocabulary through conversation.
I was in the auto parts store and a little boy saw me bring in a battery. He said "that wuks heavy!". I replied "that does LOOK heavy". Such a correction is abnormal from adult to adult. We have more grace depending on age.
For me, an average adult, with a busy schedule, learning the difference between the English T and the Spanish T made some great headway in my language hobby. Moving onward in that direction it has helped me understand spoken language - further aiding my memorization of the actual vocabulary.
My "grain" if you will:
Phonetics - Vocabulary (Anki with imagery) - Conversation (with natives, when I'm able to, it's a hobby)
The current, modern, and typical grain if you will
There is no "modern grain". Our brains haven't changed in any fundamental way over the past 10,000 years of civilization. Humans have been learning language as children the same way, for the entirety of history. The modern method of teaching language goes against the way humans naturally learn languages.
Look, talk to anyone who learned a language in school. None of them are anywhere near as good as a native speaker.
He said "that wuks heavy!"
I wonder how often children who make mistakes like this either have a hearing impairment or have parents who think it's "funny" to imitate their children's mispronunciations. Baby talk is a widespread thing and I think it's harmful to children.
When I talk to kids I treat them like an adult. I try really hard to speak the same way to everyone. If someone doesn't understand, I generally pick it up from their body language and go over what I'm saying in more detail. I ask questions to test their understanding.
> Humans have been learning language as children the same way
But you can't disregard the fact that an adult's brain is different from a child's brain. An adult's brain has a lot of prior knowledge that a child doesn't possess. Concepts such as freedom, money, market are easily grasped by an adult whereas a child has to focus on the meaning rather than the words (the representations).
I think adults can take advantage of their prior knowledge to speed up language acquisition.
Those approaches aren't mutually exclusive. You can easily learn theorems and math concepts using Anki and then apply them in practice.
You can learn words in foreign language using SRS and then use those words in conversations, when reading, watching a movie, etc.
I don't see any value in the purist approach your propose.
Also:
> So when we use this sort of spaced repetition we're trying to artificially "warm up" the cache in our brains.
No idea what this means. It is no more artificial than actively seeking out people to talk to just because you want to learn a new language, or forcing yourself to prove theorems just to better remember a mathematical concept.
I get what you mean and I agree to most of your points. I just want to point out that memorizing something you don’t use often is not completely useless. Some valuable creativity only happens when you connect some obscure and unrelated facts together, or some specific details from another field with what you’re currently working on.
> In the long run, if your knowledge is situated within a context and connected to other things you know (because you learned by doing) then I believe you'll understand it much better. Memorizing by rote, out of context, will help you to recall facts which often don't have any meaning to you.
I feel it's worth mentioning that the creator of SuperMemo agrees with you here. "Learn before you memorize" (Rule 2 in [0]). Learn the information, then commit portions to an SRS tool to keep them more effectively memorized and recallable.
You should be exploiting hacks to make things easier, not purposefully avoiding them to make things more difficult for you out of some misguided attempt at “purism.”
What second languages have you applied this technique to and how many hours of learning did it take for you to approach reading ability, conversational ability, reading/writing fluency, conversational fluency?
I want to learn a language but I don't think I'll be using it with anybody anytime soon. But I'd be happy to understand a newspaper or Wikipedia. What would be the recommended way to learn vocabulary if Anki is not an option?
I've been using Duolingo but it really feels like a glorified Anki
Content! Find video (movies, series), audio (-books, radio, podcasts).
The best way is to get exposed to the language as much as possible (some people prefer humans for this, some people prefer multimedia), and then you can use Anki on the side.
For example, I'm recently watching Fullmetal Alchemist in Japanese. I don't speak Japanese, but I've picked up 50 or so phrases. Sometimes I get a phrase stuck in my head without knowing what it means, or I'll wonder how to say something (that I know I've heard but forgot), and later that day my brain will dig up the answer from the depths (essentially a much longer version of the 'active recall' that Anki uses).
I'm a big fan of the Input Hypothesis (Stephen Krashen) which says the best way to learn a language is "input before output", in other words, immerse yourself in the language for a long period (say, 6-12 months) before trying to produce output (writing or speaking).
In this way, by the time you speak or write, you will know what it sounds like, how the grammar works, and will be able to self-correct, rather than forming bad habits.
We all know people who've lived in a country for decades and their grammar & pronunciation is awful despite constant use. The argument here is that they were forced to produce output (speak) before they were ready (got enough input to develop a workable model of the language), and were not corrected, typically out of politeness.
The key point is that the brain does this modeling on its own, given the input. Explicit learning (eg. Anki) is unnecessary, but it does "cement" knowledge and accelerates things.
If you'd like to read more on this, AJATT (All Japanese All the Time) is a great blog (a bit rambly though) on this method and also goes into spaced repetition quite a bit.
On the other end of the spectrum you have the exact opposite approach, traveling and going out to pubs and speaking the language from day 1: Benny Lewis, Fluent in 3 Months
Start with basic books but try to move up quickly to more challenging material. I read through Lord of the Rings in second grade. It was slow and frustrating at times, but the fact that the book is so engaging helped a lot. I learned so much vocabulary in such a short time this way.
Don't be afraid to keep a hard-copy dictionary of the language you're studying handy.
Been using Anki for a couple of years to learn Spanish. Was "ok".
I now use Anki to learn Spanish but now I add imagery (also known as a mnemonic). Our minds don't remember text too well - but coupled with imagery they do. To boot, Anki supports audio files to even further ingrain the memory if you have the time to create the cards.
If you're super imaginative, like the Russian Mnemonist Solomon Shereshevsky, (article on front page HN today) you wouldn't need to this do, you can create anything in your mind and recall it.
This can in way (it's physically limited whereas Anki is not) be taken to pen and paper as well using the Leitner system, a few file folders and some paper. https://en.wikipedia.org/wiki/Leitner_system
So whatever you're trying to memorize - math, names, bible verses, numbers, languages -with imagery and with spaced repetition.
I bought a book a while ago called "Fluent Forever" (not associated). The author has released an app to learn languages now using this imagery method. I personally don't like the app as it takes away the flashcard creation process reducing your "learning experience" to just a few taps. Anki, Supermemo, -insert SRS program here-, albeit harder, it's much more rewarding.
I use Anki to learn Japanese vocabulary. I have found it much more useful when I use Kanji vs. Romanization. I see Kanji differently than I do Romanization and I wonder if there isn’t a picture quality to it that gets lost when I Romanize the words?
I think there's something there for sure. The ability to rewrite the Kanji correctly probably wouldn't be there (I believe it follows a strict pattern?). I was messing around with Arabic for a month and was able to memorize written words and what they meant but could not reproduce it if I wanted to.
Spaced repetition is misunderstood. The original idea was to decrease the amount of time needed for long term memorization, over a different approach, like sequential or purely random repetition. However, many people now compare SRS over no repetition at all.
One big problem of SRS is that it gets unwieldy with larger chunks of information. SRS is good for vocabulary, bad for medical texts. The much bigger problem is that it takes an unpractical amount of calendar time. If you have three months, yes you can SRS a few hundred words. If you have a week and want full recollection at the end, you'll still repeat the list of words a couple dozen times and SRS makes no difference.
Another problem is the time it takes to prepare a database or a set of paper cards. By the time I prepare a database from a book I could probably read the thing twice or more times.
For exams that require recital of information (for example medical or law exams) mnemonic techniques like the Loci method (also called memory palace) are much more reliable and more practical.
I’m OP. Thank you for the feedback!
I don't see why not comparing "SRS over no repetition at all". To me, SRS or no SRS, they are all learning/studying methods.
All SRS algorithm is expanding, i.e. increasing intervals. So maybe with continue success, it'll be: 1 day, 2, 4, 16, 40, 300 (arbitrary numbers).
How do you propose to lean medical texts, if without SRS?
Mnemonic techniques only work with rote recall, whereas flashcards, if formulated well, can help induction, understanding, not just recall.
The idea with SRS is that the exponential scheduling is supposed to be more efficient than random or sequential scheduling.
I did study for medical exams with mnemonic techniques. Studies have shown that they do not inhibit understanding at all. In my experience they even help understanding. They totally dominate SRS on anything looking like a table or list. With text it's not as pronounced, but still you can memorize the key words, as few as possible, as many as necessary.
Flashcards are first of all a rewrite of the original material. If that is really necessary, the original material sucks, at least for the purposes of the exam. The physical cards are marginally easier to use than a textbook, but then again, writing them takes a lot of time and effort.
Usually when people talk about SRS, they mean flashcards or Software. At the very least you would need to keep track of the learning schedule and what you forget. Which I found impractical for learning large amounts of stuff.
I'm a medical student who uses Anki religiously. I'll vouch for it any day. Currently, I see around 300-1,000 flashcards a day and have been doing it for almost a year straight. Without Anki, I am not sure if I'd be doing nearly as well in school. It is really a game changer for my education (at least my medical education). I wish it was more user friendly, but it is free (on Android) so I can't complain.
As for Supermemo comparison, Anki has been much easier for me to sync / utilize decks offline. I just keep a 24K card deck on my phone and sync it occasionally for a backup then I'm good to go.
I've had the idea that a lot of medical training (besides anatomy, chemistry, bio, many other things :-)) was memorizing a big decision tree to evaluate someone, which seems fundamentally different than flash cards. How do you learn that? Family history that pertinent , previous conditions etc.
Hi Latteland! I do both. Anki is about the little facts that I want to retain after I understand the bigger picture. In fact, I never use Anki to learn anything and only for retention.
Example flashcard:
- Front: Which race has a higher prevalence of G6PD Deficiency?
I’m OP. Thank you for the feedback.
Yes I have no affiliation with SuperMemo whatsoever. My previous 5-year Anki experience gives me in a unique position when comparing both software.
A few months ago I built a Chrome extension that replaces the new tab page with an anki window, in an attempt to improve my browsing habits and it seems people overall are finding it useful so far,
I spend an insane numbers of hours on airplanes with my iPad. Any recommendations for which of these would be good for creating cards (with Apple pencil) and using on an iPad with no Internet connection?
The main hump I need to get over before committing to either tool is deciding what's important enough to remember. If I were learning a language, using spaced repetition would be obvious. In most other areas, it's not so clear whether it's important to remember the contents of any given book or article.
I suppose I could apply it to remembering songs I've learned how to play, but I'm just doing it for fun.
There's an actual program, and a saas version. The saas version isn't worth paying for (imo) over using AnkiWeb unless you wanted preformatted decks, and the most up to date version of the algorithm.
I've experimented with using the website (mostly because I'm a fan of the software). It's mostly okay. I've been using Supermemo (the actual app) for a few years now on a daily basis. The app itself has a trial if you can find a download link somewhere, and I believe the site might be free if you import your own decks.
As far as lock in, I believe you are also locked in unfortunately. This would be another reason to go with Anki or another solution if you just want flashcards.
We are talking about https://super-memo.com/. The supermemo.com/SuperMemo World is different. I never used the latter. I think they are managed by two different groups of people.
SuperMemo 15 is free. The latest version is paid software unfortunately.
The actual desktop SuperMemo app (supermemo 17 I believe) is not subscription based and is way way more powerful (but also convoluted and somewhat outdated UX wise). The web supermemo is dumb down and mainly for language learning.
I am an Anki user and would love to try SM but the UI is so clunky it is so hard to use. I wish someone put together a comprehensive guide/book on using SM. I went through a few on YouTube from the creator of SM and they were okay but it was still hard to use.
Supermemo web doesn't let you export your cards. Maybe there's a way to do it, but I'm using anki so I never lose my cards in case the service goes away or has data loss
There is most certainly a way to do 'incremental reading' in Anki, though it must be done manually. You can do virtually any type of reviewing in Anki with its high degree of customization.
You can import entire articles into Supermemo and break them down into flashcards without leaving the program. Anki just gives you flashcards. There's an incremental reading plugin for Anki, but it doesn't work very well.
Incremental reading works by breaking down an article into flashcards. Basically, you go: article -> paragraph -> sentences -> flashcards (usually with cloze deletions). But, because the algo manages when you see everything, I don't have to worry about doing all of that in one step.
Polar Bookshelf (getpolarized.io) is more of an open source solution for archiving documents at the moment. You can export annotations/extracts from Polar into Anki, but you don't get the spaced repetition with the actual reading of articles like you do with Supermemo.
I read a bit of the article, decide if making an extract of that portion is worth it, and then click next repetition when I'm done/bored reading it. I'll then get another article/flashcard/extract (of another article).
You also have tools for marking up articles and saving your place.
Every day I review around 30-50 flashcards, and 30-50 articles/extracts of articles.
With Supermemo, I don't have to decide (or rather, I give up the ability to decide) what I'm going to read next; in favor of having everything scheduled by the algo.
I go through about 800-1000 items in my RSS feeds once a day, save them into read it later/polar/bookmarks, and once a week, I throw the ones that still seem interesting into Supermemo.
I probably spend 30min to an hour going through everything once a day, and another hour or so on Supermemo. Most of the items in my RSS feeds are admittedly junk around 95-98% or so (HN and HN comments, reddit rss feeds and comments, lobste.rs and comments, some news sites, a few personal blogs, and some forums), but I do get a few gems throughout the week.
The newer Supermemo algorithm also tolerates breaks better than Anki (the last time I checked). So, for example, not using Supermemo while on holiday isn't an issue.
I also keep aha moments from programming and some documentation in there as well.
Edit:
Here's a demo of incremental reading in Supermemo. He marks up articles much more than I typically do. There's a lot of features and functionality in the program.
I'm a bit surprised these systems still pay off for people. How many contexts require learning large bodies of unrelated facts, these days? Everything's just a google away. Even in the past, the only context where I would have used a system like this is language acquisition, and it's pretty clear that machine translation is going to surpass me very soon, if it hasn't already.
For me, most interesting work involves conceptually coherent bodies of knowledge. Learn the conceptual framework, and you'll remember the details by partial reconstruction as you need them. At least, that's how it usually works for me.
All of them except programming, I'd say. That sounds like a 2000's-programmer-centric view of the world. Modern software is unique among all the fields I've worked in, in that it combines:
- people routinely working on systems far too complex for them to ever understand fully, and which often change completely on the timescale of weeks or months
- all the relevant data on the web in a format (like error messages or function names) that's easy for google to index, and easy for users to type
- people sitting in front of internet-connected computers anyway, and sufficient time slop that the team doesn't need an instantaneous answer
When I was knee-deep in water at an archaeological dig 15 years ago and my professor called out "What species of tree is that by your knee?", or when I was working on the catwalk last week and my lead called over "Check that the instrument in front of you is safe", googling it would have been completely infeasible, for several reasons. They'd just replace me with someone who knows what they're doing.
> Learn the conceptual framework, and you'll remember the details by partial reconstruction as you need them.
When I took organic chemistry, they taught the special cases of each reaction type first, and then generalized them at the end of the year. When I took engineering physics, they made us start with the generalized form first, and then we had to derive the specifics as needed. At the time, I was frustrated by the chemistry approach (why didn't you tell us there was a simpler form?!), but 20 years later, I remember a lot more chemistry than physics (even though my grades in physics were much better). I might be able to fumble through an SN2 reaction if I needed to, but I'm sure I couldn't do anything at all useful with Maxwell's Equations.
Medicine is a field I'm in where learning large bodies of unrelated facts is helpful, since it's a field where having a generalist's shallow understanding of a wide variety of fields is useful.
Take biochemistry, for example, wherein one can see it as a series of mechanisms and is mostly memory work. Anatomy is similarly mostly memory work; one doesn't simply "reproduce" an understanding of the course of the vagus nerve from scratch.
Unfortunately, high schools and university courses can rely on this sort of memorization quite heavily as a proxy for "understanding". My biochemistry courses, particularly early on, were at least 60% memorization, 40% conceptual/logical reasoning.
Spaced repetition is a very useful tool when making it through law school; it was for me and many of my cohort. It's very useful to improve scores on many tests, from first aid to tests for a hunting license. Is it very useful to 'really' learn? Only a bit. But there is value in improving test scores in its own right.
I find it useful to re-solidify the bits pf a conceptually coherent body of knowledge.
I’m also going to soon need to work on a knarly php codebase for what are some fairly compelling business reasons. I’m planning on making flashcards for the body of unrelated facts that is the php developer experience.
I skimmed this entire article out of curiosity. It uses the term SRS twenty times but never says what it means. I went back to the top and noticed the "not a prerequisite" link, so I looked at that article too. About six paragraphs in, the fourth mention of SRS finally gave me the key: "Spaced Repetition Software (SRS)".
This tiny bit of knowledge gave me the context to understand what the articles were talking about.
If you put definitions like this at the top of every article, you will have more informed readers who can better appreciate your insights.