Here Is Why Most Spaced Repetition Apps Don’t Work and How to Fix It
Regardless of whether you use Spaced Repetition Apps or not, you can’t deny that there is some controversy among language learners whether such programs are truly effective. Some people swear by it while others prefer more old-fashioned pen-centered strategies. It gets even better! Even among SRS enthusiasts, you can find different militant fractions. Some claim that Memrise is the best. Other that Quizlet is the way to go.
For many, it can be quite difficult to wrap their head around what’s true and what’s not. Let’s sort it out so you can finally know the answer.
What’s the scientific consensus about Spaced Repetition Apps
If you have ever seen one of the aforementioned squabbles online, the first thing you need to know is that opinions that SRS is ineffective are completely detached from reality. Spaced repetition is among the most thoroughly researched memory-related phenomena in the world. Its efficacy has been replicated in hundreds of comprehensive and extensive studies (read more about choosing the best language learning methods).
It is effective on a variety of academic fields and mediums.
Spacing effects can be found in:
- various domains (e.g., learning perceptual motor tasks or learning lists of words) such as spatial44
- across species (e.g., rats, pigeons, and humans [or flies or bumblebees, and sea slugs, Carew et al 1972 & Sutton et al 2002])
- across age groups [infancy, childhood, adulthood, the elderly] and individuals with different memory impairments
- and across retention intervals of seconds [to days] to months (we have already seen studies using years)
Source (probably the best article online about the spaced repetition, well worth checking out)
The benefits of spaced study had been apparent in an array of motor learning tasks, including:
- maze learning (Culler 1912)
- typewriting (Pyle 1915)
- archery (Lashley 1915)
- javelin throwing (Murphy 1916; see Ruch 1928, for a larger review of the motor learning tasks which reap benefits from spacing; see also Moss 1996, for a more recent review of motor learning tasks).
Heck, there are almost no exceptions to this phenomenon. Sure, there is maybe 5% of studies which haven’t replicated these findings. But upon reading more about their design and methodologies used, one might conclude that they are often an example of bad science.
The only notable exception I have seen so far is that children can often fail to exhibit a spacing effect unless they process learning material in a certain way. This, however, is a topic for another article.
Where does all this controversy about the effectiveness of SRS programs come from then? I will get to it soon.
First, let’s concentrate on what makes learning truly fast and effective.
Encoding – the most important criterion for effective learning
A simple model of memory
The process of memorization can be depicted in the four following steps:
- Retention intention
- Encoding – involves initial processing of information which leads to the construction of its
mental representation in memory
- Storage – is the retention of encoded information in the short-term or long-term memory
- Recall – is the retrieval of stored information from memory
Let’s concentrate on the second step of this process. Clearly, you can see that it’s a gateway to the land of remembering. But what does encoding really mean?
“Encoding is any kind of attempt of manipulating a piece of information in order to increase your chances of memorizing it.”
What’s more, there are two kinds of encoding.
Two types of encoding
Shallow encoding doesn’t help you to connect the piece of information with other meaningful information nor does it help you to further your understanding of it. It usually concentrates on meaningless banalities.
Example: you are trying to memorize the word “skada” (Swedish for “to damage”). The prime example of shallow encoding would be to start counting the number of vowels or consonants in this word.
The absolute opposite of shallow encoding. This time you are trying to make a meaningful connection between different items. The more the better.
Deep encoding is so powerful for your learning that it even shows up in brain scans as increased activity in key brain areas associated with memory. It is this activity that appears to give deep processing its memory advantage. (source: How Memory Works–and How to Make It Work for You).
So what’s the example of deep encoding in the world of language learning? Creating sentences or saying them out loud, to be more precise.
Interestingly, every time I say it, there is always someone who seems surprised. I guess the reason being that we don’t appreciate enough how complicated it is for our brains to create a sentence.
Why creating sentences is so complicated
In order to create even the simplest of sentences you have to:
- remember actively the words you are currently learning
- remember all the other words in the sentence actively
- connect them in a meaningful way
- apply all the known grammar rules
- choose the appropriate register of the sentences (i.e. a form of a language used for a particular purpose or in a particular social setting)
- remember the pronunciation of all the words in the sentences
- pronounce all the said words by using your muscles
As you can see, it’s not that trivial to produce a sentence. And that’s why this process is so meaningful and memorable for your brain.
Initially, a lot of my students grumble about having to create many sentences. They say it’s too exhausting. I agree. The thing is that producing sentences equals knowing and being able to use a language!
To make your inner geek happy, it’s worth mentioning that encoding is very often connected with two other principles of memory which make your learning even more effective:
The level of processing effect (Craik & Lockhart, 1972) – the more you process a given piece of information, the better you remember it.
The generation effect (Slamecka & Graf, 1978) – active production of a given piece of information increases your chances of permanently storing it in your long-term memory.
Read more about optimizing your language learning here.
Interesting, right? Now it’s time to answer the most important question – what if somebody is too lazy to actually go through all the trouble of producing sentences?
Consequences Of Lack Of Encoding (i.e. why most Spaced Repetition Apps don’t work)
I hope that the following paragraph will help you make a very important decision – never ever use or buy any learning app. I don’t care that you read that Gabriel Wyner is working on a revolutionary app or that Memrise has a better algorithm now.
The most important and effective thing you can do for your learning is to create multiple contexts (i.e. sentences) for a word you want to learn. Simply repeating ready-to-use flashcards, especially the ones without any context, won’t work well. This simple fact renders all the memory apps combined useless. ANKI is really all you need.
Think for a second about the solution those apps dish out to you. Most of the time they simply give you ready-to-use flashcards, often without any context! Or meaningless games which perpetuate shallow encoding. Or even when you see a flashcard with a word in the context, it was not encoded by you and thus it will be way harder to remember.
Time to stop looking for magical solutions. You won’t find them in apps.
To my chagrin, I don’t see any big company talking about this. Of course, the reason is obvious. If you pay for an app, you have to be convinced that it’s truly magical and life-changing. I don’t think they would sell well if the owners started screaming from the rooftops “They are sh*t! What’s truly magical is the effort you put into encoding your vocabulary”!
Read more about Common Language Learning Mistakes and How To Fix Them With Lean Language Learning.
SRS programs are just a white canvas
The right way of thinking about such programs is seeing them as a white canvas.
Algorithms underpinning them are close to perfect in themselves. Unfortunately, some people crap in their hand and insist on smearing it until they get a one-eyed unicorn. The next thing you know is they are running around the internet and screaming that SRS programs don’t work. You can’t be lazy when you learn.
I know that doing ready-to-use flashcards seems “quicker” to use because you don’t have to invest too much energy into producing them. However, in reality, they are more time-consuming in the long run because you need to spend more time repeating words unnecessarily.
It has to do with the mechanism of passive rehearsal which is simply a mindless act of rattling off a cluster of pre-prepared information. Many years ago it was actually proven that it has little effect on whether or not information is later recalled from the long-term memory (Craik & Watkins, 1973).
If you ever want to use such flashcards, simply treat them as a source of vocabulary to learn. Other than that, simply encode your vocabulary and you will be fine. All ready-to-use flashcards can do is create the illusion of time-efficiency while slowing your progress down at the same time.
To sum up, currently there is no other technology, including virtual reality, which is as effective as spaced repetition programs. However, if you don’t actually put in the effort and try to produce sentences for the words you learn then you waste most of the potential of this software.
Quick learning is not about time but about the effort.
Done reading? Time to learn!
Reading articles online is a great way to expand your knowledge. However, the sad thing is that after barely 1 day, we tend to forget most of the things we have read.
I am on the mission to change it. I have created over 30 flashcards that you can download to truly learn information from this article. It’s enough to download ANKI, and you’re good to go.
When making your own flashcards, how do you ensure that the sentences sound natural in the target language? I have been hearing lately that I say super unnatural things in Japanese sometimes.
Even when I make my own flashcards and use sentences I’ve remembered in real-life scenarios with people and think that genuine experience will make them Really stick… I still forget the words when the card comes up for review again. Even though I use the “hard” button on Anki to put reviews closer together 90% if the time. What gives? Anyone else struggling with this? It’s like the words are falling out of my longterm memory.
Wonderful articles! And I’m amazed at your thoughtful and detailed responses to those who leave you comments!
Do you have an article with your recommended process of optimized learning in ANKI? Do you recommend seeing a word in your native language on the front of the card, and then on the back of the card, the sentence you created in the target language using the translated word? If so, might it also be helpful to have the card be a “type in the answer” card, where after seeing the word on the front of the card you speak out loud the sentence you’ve memorized and then type it in to continue to enforce the original encoding with both speaking and writing? Thanks in advance! 🙂
Thank you! No, I am afraid not. I leave most of the methods for my courses 🙂 Best! Bartosz
I feel much better now. I used to lament that I had to learn Japanese in the pre-Internet era of the 1980s, which I’ve claimed took longer to accomplish because of the limited resources. However, I did manage to memorize 2,500 Japanese (really Chinese) characters within an amazingly short period by creating flash cards on physical index cards and then drilling them afterwards. Even then, I recognized that I seemed to have learn them mostly by creating the cards rather than while reviewing them. Now, I realize that why this was true. I’m glad that I did go to the trouble to create those cards. It’s likely that I wouldn’t have learned the characters nearly as well if I’d purchased a box of ready-made flashcards. Now, years after having discontinued my career as a technical Japanese translator, I’m trying to learn 11,092 Chinese words with various flashcard sets that I’ve created and arranged on Quizlet according to different criteria. Whenever I find it difficult to remember the correct reading of any particular Chinese word that doesn’t happen to exist in Japanese, I look up the meanings in English, Japanese, French, and sometimes Swedish to give me a different perspective, including native Chinese dictionaries as well. Later, the pronunciation comes more readily to mind even though the focus of my rea search was more on meaning instead of pronunciation just because I’d spent more time thinking of the word from various perspectives. Based on how I learned Japanese, I still think that learning the pronunciations with the correct tones for these 11,092 words is still the best first step in my case, though it’s far less important nowadays when we can copy and paste characters into dictionaries without having to know their pronunciations beforehand. It’ll take a few weeks, and then I’ll move to my second step of learning the English and Japanese meanings of these Chinese words, which is going to be very labour intensive (creating the individual Quizlet entries for meaning). Of course, I have the advantage of knowing the meanings already of the thousands of individual characters because of their use in Japanese, which fortunately also shares many of the actual words that I’ve been using since the 1980s. It helps also that I did study Mandarin at university for three years back then, was married to a Mainland Chinese for 24 years, and have been surrounded by mostly Chinese speakers and signage in this part of Metro Vancouver since 2003.
This article is interesting, I’ve been reading “Make it Stick: The science of successful learning”, and the authors there-of discuss related phenomena: essentially, if you have somebody learn a piece of information and test them on it, the more work it requires to complete the test (up to the point where one can still successfully complete the test), the better the information is retained.
I’m a bit confused as to the *specific* method you recommend employing to create review cards, though. Are suggesting that one should write their own sentence cards for anki, then review those cards? Or are you suggesting that one should write a simple vocabulary card (target language front, definition and grammar notes on the back) then, for review, practice producing a sentence that uses that word? So each time that word comes up, produce a sentence that uses the word, then check the grammar notes / definition to see if you used it appropriately?
Hello Fred! Thank you for your comment. Yes, I am suggesting to write your own sentences for ANKI and then review them. Although, this dry sentence doesn’t even begin to cover the entire process of optimized learning in ANKI.
There are lots of other memory principles that one should keep in mind in order to truly learn effectively. For that reason, I wouldn’t recommend the flashcard format you have mentioned in your comment.
Wonderful article! So great to hear you promoting deeper methods of learning than cramming flashcards.
>>> when you see a flashcard with a word in the context, it was not encoded by you and thus it will be way harder to remember… To my chagrin, I don’t see any big company talking about this.
My company (Dendro) is not big, but we’re developing tools to specifically support people to do exactly what you’re talking about: encode memories themselves! We also include flashcard tools, but the effect is totally different when learners have thought about content and spent time formulating/encoding ideas into a (personally) meaningful form.
Hi George! Thank you! I hope you will come with something better than the existing solutions. Good luck! 🙂
Hi Bartosz! Great article!
I am currently stuck between creating my own physical SRS-system using flash cards, and “quizzes” for myself. However, I have also dowloaded ANKI and some decks created by others, but haven’t used it because I am a little apprehensive about using a digital format and handwriting things tends to be “my way.” Have you noticed a difference in the use of ANKI (either one’s own created deck, or other people’s decks) as compared to creating one’s own physical decks (beyond the difference in how much physical space is being used to store the decks)?
Hi Kaylee! Thank you! 🙂 Longhand has only a very limited advantage over typing which disappears after a couple of weeks. You can read more about it here: https://universeofmemory.com/writing-or-speaking/.
The only difference I noticed that physical flashcards are terribly limiting. I don’t want to discourage you from “your way”, but you can’t possibly optimize your reviews physically in an efficient way. What’s more, the more and faster you want to learn, the more self-limiting physical flashcards become. You can’t beat algorithms. Of course, if you don’t care about the pace and you get those happy feelings when you use physical flashcards, then you can stick with them 🙂
Thank you for your response and the link, Bartosz 🙂
It seems as though through reading a few more of your articles and checking out the source material associated, I’ve made up my mind, and ANKI is the way to go because of the efficiency and portability of the tool.
Keep up the great work!!
I am happy to hear it! Thank you and good luck to you! 🙂
great find, your place i mean Bartosz
a question or 2 if you please if you please
first, if your comments apply only to language study? you see there are many definitions say in law, or even more in medicine, from which one cannot or would not dare vary. by encoding do you mean you are the one to write the term into the app? that you should not download or copy-paste these terms
also, when a term or idea comes up, say the idea of unjust enrichment, i ask students to make an example – say you go on an emergency errand that requires you to be out of town for 5 days and forgot about your 5 huge rottweilers and i feed them for you, that creates an obligation to reimburse.
it would be great that you are able to create your own example but my question goes to whether downloading stuff in law or medicine to one’s app for study would compensate???
(1) No, the content of the articles applies to any field of knowledge. Encoding means manipulating knowledge in some way, reshaping it or applying it. I teach doctors on a daily basis, and this is the same assumption that they always have. They think that if you have to learn verbatim what’s in front of you, there can’t be any deep encoding. That’s wrong.
What about applying your knowledge? If you read that vitamin A and D enhance zinc absorption, then you can start creating realistic scenarios to use this knowledge. For example, if you have a patient with really bad acne and irregular menstruation, then you can potentially assume that they are zinc deficient. In this case, your flashcard can look like this:
Q: Patient with acne and irregular menstruation – what vitamins can I prescribe to fix these problems?
A: A and D – they enhance zinc absorption
Of course, there are lots of techniques that might be applied here but they would take too long to explain.
(2) Yes, your concept of creating examples is a great idea.
Encoding is absolutely necessary, it is the personal context in long term memory that connects to new information.
Students need to write their own definitions in terms and context that is familiar to them.
Consider encoding as a clue, a connection between old and new. A clue that is in each students context — include a clue in your flashcards — “How did you connect this new information to something you already know?”
I would like to see a flashcard program that includes Spaced Repetition and personal clues. Both would strengthen learning and recall.
I absolutely agree, Great observations. Thank you for your comment, Dennis!
Hey Bartosz, great article!
I would say that another reason why people frown on the use of space repetition apps is because they can get very taxing and boring (it’s significantly easier to do a lot of reading/listening/watching) and the way people attempt to ameliorate these issues renders the software ineffective (lessening active study, especially encoding).
Regarding vocabulary acquisition using Anki, I have some questions:
1) When learning words with multiple meanings, do you advise on actively reviewing all of these meanings?
2) If so, wouldn’t keeping up with the number of reviews get unfeasible in the long run? Would activating just the main meaning and leaving the rest to passive study still be effective?
3) Is there a point (number of reviews?) in which active study could be ceased? Because, for instance, in our native languages, there are a lot of obscure words that we rarely use/see and yet we can easily recall them if necessary, does this happens because have reached a certain number of uses/reviews?
Sorry for bothering you with the many questions!
I agree. It can get boring but as long as it’s effective, I am ok with it.
(1) Every meaning = a separate flashcard. Unless you build cues (i.e., triggers) for every meaning, you will have a hard time recalling it
(2) In the long run, everything becomes more feasible. After all, you have more time and fewer reviews each day. Yes, activating the main meaning could work if you spend enough time absorbing your target language passively
(3) Yes, I believe a C1 level allows you to cease active learning. At this level, you are expected to understand 99% of everything you hear and read, so learning more without any specific goal becomes a bit counterproductive.
What is your opinion on supermemo.pl?
I like their programs very much and I used them in the past when I didn’t know any better. Right now, I wouldn’t bother for the reasons mentioned in this article.
I would prefer to find/spend my money on a decent frequency list and start knocking out sentences on my own.
Could you do an article (or link to articles) on encoding? This has always been my greatest struggle, even in mnemonics. Like how do one systematically encode knowledge when there is no way to know upfront what form it’ll take? I always end up having to come up with a new to encode things I come across and it becomes messy and tends to be time consuming and ineffective. :/
That’s a great idea! Thank you for your suggestion!
I’d be interested in opinions on https://www.kartoj.com/ which is a spaced-repetition mobile web app for language learning, using the Fluent Forever philosophy. (Currently supporting Spanish learning, but other languages will follow.)
Nothing special. No real context present. I can’t access the words I learn. I definitely wouldn’t use it. Pictures are nice but not as useful as people think: https://universeofmemory.com/pictures-and-images-in-flashcards/
Hi Bartosz, fantastic site!
I have two questions:
1.) When you recommend doing own flashcards based on frequency vocabulary lists – isn’t there a risk, that complete beginner would create sentences, which are completely unnatural or simply rubbish? Whenever I did any card, I always relied on WordReference.com or similar sources with examples created by somebody else. Is Assimil or pre-created decks from Anki users really step in wrong direction at the beginning of learning process? I realize that you have your course related know-how and cannot reveal everything in detail, but I am still somehow obsessed that I am doing everything wrong using aforementioned resources 🙂
I would like to notice, that somehow I still remember sentences studied this way in Anki, another question is long-term retention and ability to use such vocabulary in meaningful conversation, which I have not tested yet.
2.) I looked for some scientific papers dealing with method based on memorizing whole excerpts of text, almost everything truly relevant is inaccessible for free, have you ever tried this? I have found one study where were mentioned students who learnt this way, they even won some debate contest in foreign language (I guess it was English) and referred to this memorizing as a way they had built their fluency. I would appreciate your opinion.
Thank you for your inspiring and truly informative work!
Hi Martin! Thank you!
1) Of course, there is. It’s inevitable that initially most of these sentences will be terrible. Effective learning is not about being right from the very beginning. It’s about automating your knowledge. So what that you make 5 mistakes in one sentence. The only thing that matters is that you remembered about 1 grammatical construction and you used it. That increases your chances of using this construction correctly again. Once you automate it, you can move on to the next one, etc.
You will always remember some sentences. The funny thing is that if you remember full sentences, your knowledge isn’t very well consolidated. Otherwise, it would be generalized – you would be able to recall a given word without any effort (just like in your native tongue) but you wouldn’t remember concrete examples.
2) I haven’t tried it. It doesn’t make much sense to me. Producing speech is a spontaneous act. In order to do it you need a lot of smaller building blocks, not whole excerpts. I am sure it makes sense when you debate as you learn the whole argument but in the real world that seems pointless.
Hope that answers your questions!:)