January 31, 2022

My random thoughts on language learning

My random thoughts on language learning 

Mastery/no blind spots

Inspired by Salman Khan, while learning a particular concept, try to learn in a way that you understand everything you are learning perfectly with no ‘blind spots’ in your understanding. In other words, if tested on what you are learning, you should get 10/10, ten times in a row. It seems, that the information you learn needs to be perfectly understood by your brain so that it can turn into a solid, automatic, functional block quickly. 

Additionally, you can’t just learn something and know how it works roughly. It’s like learning what the gas and brake pedals do, but not learning what the clutch does and then in the next lesson being taught how to parallel park. Obviously I need to know what the clutch does before I can do that. 

Law of distributed burdens*

It is wise to divide brain-intensive operations into separate, more manageable chunks. While logical, it’s unfortunately not obvious and this strategy has helped me many times deal with brain intensive operations, not only when it comes to language learning. 

For instance, it is very inefficient to learn Chinese by learning how to read, write, speak and understand the language all at the same time. If you distribute the burden, say, learn how to read in pinyin first, then move on to listening, then move on to text-chatting in pinyin, then to speaking, then to reading Chinese characters, then to learning how to type Chinese characters and finally to handwrite, the task will be much more manageable.

This approach can be applied to lower levels of language learning too. For instance, it is more efficient to try to improve your pronunciation one sound at a time rather than try to improve several sounds at once. It’s for instance more efficient to consciously accept that you are not pronouncing the Spanish ‘r’ correctly and leave the improvement of this sound until after you’ve become somewhat fluent in the language. Working on both improving the ‘r’ sound while trying to improve your general fluency at the same time might be overloading the brain and actually slowing down the progress towards fluency itself considerably.  

*I don’t have a better name for this concept unfortunately but I hope it will do. 

Possible relation between subitising and automation

Subitising is this: If there are 1 to 4 apples on the table, you don’t need to count how many apples there are, you know their amount by simply looking at them. If, however, there are 5+ apples on the table, you need to count them, to be able to tell their amount.

I’ve been thinking about this quite a bit and I feel like this concept is deeper than just knowing vs. having to count the amount of something. I feel like, maybe (I have no study to back me up), the human brain is able to work with up to four individual pieces of information at the same time but not more. 

This in turn, when trying to automate a concept (say automate the conjugation pattern of some verbs), could be key in knowing what is the maximum amount of individual information building blocks that a concept can consist of until it gets too overwhelming and optimise for that. Otherwise you just might not be able to automate it effectively.  

After participating in one study on subitising at the Humbolt University in Berlin and reading about the subject, I found it interesting that there is such a big difference between telling the amount of 1-4 objects and 5+ objects in terms of how the brain has to deal with these two operations. I also noticed that when I try to understand a concept, be it language-related or not, I can also consciously work with roughly 4 individual concepts/information blocks at the same time and thought it would make for a good topic for a healthy debate as to whether the phenomenon of subitising maybe is something deeper than only being able to tell the amount of objects instantly or not.

Whatever you are learning must be automated for it to be functional. For instance if you learn the word for “table” in German, it’s of no use when you have to spend 10 seconds remembering how to say it every time you want to say it. Automating this one word might be relatively easy, because it’s only one word, but in language learning (and learning in general) a concept we are learning often consists of several individual building blocks and that number of building blocks is often greater than 4 and your brain might be overloaded when you get past this number and switch to a different, less effective 'style of perception'. Just like having to count the apples on the table past the amount of 4. 

A classic example would be learning how to conjugate verbs in Indo-European languages. There are usually 3 persons and 2 numbers (6 information blocks) which might be too much for the brain to work with at the same time. 

Being aware of the fact that we might get overloaded and confused if we have to work with more than 4 individual information blocks to understand a concept could improve our language learning strategies. Limiting the number of individual concepts/pieces of information to be automated to four might lead to more effective learning.

Just in general, there obviously is a limit to how many individual pieces of information we can work with in order to automate a concept, so one next question logically is, what that limit is. Looking at the similarity between subitising and general concept understanding, I propose, the limit might be 4. 

It's only an idea of course with no study to rely on, but I thought the concept was interesting.

The influence of tribal psychology on language learning

I don’t have any controlled double-blinded study to back me up on this either but based on experience and discussions with my friends, the subconscious willingness to be or to not be a member of a certain ‘tribe’ seems to be an important factor influencing the rate at which you improve in a certain language.

These might be emotionally charged words, but if you really love a certain culture/nation, compared to when you’re learning a language of a culture that you don’t really want to associate or identify with your learning process should be much faster. Again, I have no study to back me up on this, but I feel that the process might even be different on a neuron-to-neuron level.

If, for instance, you love Italian culture, subconsciously you would be more willing to sound like an Italian, imitate the Italian accent and mannerisms because you are willing to become part of the ‘Italian tribe’ and on a neuron-to-neuron level your learning might be much faster. 

If, on the other hand, you don’t feel like this about, say, the Chinese culture, tribal psychology might be the reason why your Chinese isn’t improving very much.


It sounds like a cliche at this point, but passion is the holy grail of learning, not just language learning, in my opinion. Apart from the fact that a passionate learner will actively try to improve not only the language he or she is learning, that learner will also actively seek to improve the way that person is learning. The old saying “languages cannot be taught, they can only be learned” really hits home here, because a passionate learner is actively seeking to learn the language by default. Sometimes a student like this doesn’t have to be taught anything and will come up with small improvements and tweaks to their strategy on a daily basis, personalised to their learning style which no teacher would ever think of. 

Additionally, learning a language to proficiency is not something that can be done in a few days. It takes months and usually years, so superficial interest is not enough. Passion is what keeps the student invested in learning for a very long time. Passion also makes the student think about what they are learning every day which is basically organic spaced repetition.

The importance of passion cannot be underestimated and if learners are failing to learn, maybe they should take a step back and analyse the situation in these terms. 


In my opinion, what people refer to as talent in language learning could be broken down into:

  1. pattern recognition ability
  2. time it takes for a passive information block in your brain to become active, automated and functional 

Compared to other factors such as memory or the ability to imitate sounds, the above two, I feel like, are innate and cannot be trained or improved on by much in our lives and therefore if there is such a thing as ‘talent', which is a unique biological predisposition of every person, it might be this. 

On a side note, I feel like pattern recognition is just a sign of general intelligence and thus a person generally intelligent and generally good at learning things should also be a good language learner.

The ‘automation’ of information however, I do feel, is an inborn quality like blue eyes and cannot be improved on by too much. Again, no scientific study to back me up on this. I’m only a philosopher at this point:)

Confidence and the Dunning-Krueger effect

One reason why children learn languages very quickly in my opinion is, because they have very little doubt that what they say is correct/perfect and what they hear they understand well/perfectly. Doubt and constant back-checking can completely destroy one’s language learning flow and is unfortunately a direct consequence of growing up. With experience people learn that there is often more to things that seem very simple at first glance and are by default subconsciously expecting a concept or rule they are learning to have more variation and exceptions than necessary which impedes swift automation.

The Dunning Kruger Effect is a situation where the less you know about something the better you think you are at doing it. An example would be beginner learners of Mandarin Chinese saying that Mandarin Chinese grammar is easy and that the most difficult thing about learning Mandarin Chinese are Chinese characters.

Unfortunately, to this point I haven’t really figured out any mental trick how to feel confident when I’m aware of the many exceptions a certain concept or rule might have and variants the concept itself can come in. If you look at the Dunning-Krueger curve, you can see that your confidence starts reaching your initial levels once you have enough experience but, where that limit is, I don’t know:)

However, at least being aware that doubt and constant back-checking is destroying your learning flow and is harmful because it impedes automation is very important too. 

Aha! moment Observation horizon

In black holes, the ‘event horizon’ is the place past which you cannot see so I was inspired by this to call the place past which you simply cannot observe a certain concept because you don’t have enough information, the Observation horizon.

For instance, you can't understand Calculus, if you don't understand simple arithmetical operations, algebra, how the Cartesian coordinate plane works, what limits are and so on. You can't 'observe' the concept of Calculus, not because you are not smart enough, but simply because you lack the information building blocks that build up Calculus and you are thus too far away from the Aha moment Observation horizon for Calculus.

Same applies to language learning. Whether small individual rules or understanding the language as a whole, it is helpful to realise how far you are from the Observation horizon. As mentioned above, automation is key, automation can only happen when you understand something, preferably perfectly and that can only happen when you can observe that something.

On a side note, I was thinking that maybe a scientific experiment could be set up to measure general intelligence/pattern recognition this way. A picture of a cat (outlines only) would start gradually randomly revealing itself in 1% increments, and people tested would have to stop the process once they would realise that it is a picture of a cat. The less of the picture revealed would mean, the person was able to recognise the image was an image of a cat while being further away from the Observation horizon, has greater pattern recognition abilities and greater general intelligence. 

The great minds of the past, Einstein, Euler, Newton, were different in that they were able to notice a certain phenomenon with less information about it revealed than their peers. They were able to identify it being far away from the observation horizon. Today you can find dozens of young guys on YouTube confidently explaining to you the theory of relativity thinking they’re very smart because they can understand it, but that’s because they are standing right on the observation horizon. All information building blocks necessary to observe the phenomenon of the ‘theory of relativity’ have been already discovered and all one has to do is familiarise oneself with them. I would however like to see them in the times of Einstein with the amount of information blocks Einstein had to come up with the theory of relativity. 

Every intellectual problem that at least one human has solved before, meaning there is proof that it can be solved and learned by humans, you can learn or solve as well. 

For example, calculus can be daunting, but when it comes down to it, it just consists of several information blocks that have to be understood individually before you can proceed to understanding calculus as a whole and the reason you don’t understand calculus is because you either 1) didn’t take your time to understand the individual information building blocks it consists of or 2) didn’t take your time to piece them all together to understand calculus or both. 

I used to look with admiration at quantum physicists and think, How are they able to understand all this? and the answer is simple. They just put in the time to understand the individual building blocks that build up quantum physics and put them all together which I haven’t. 


  1. Great article, as usual. Even if you haven't yet found any articles to back you up on this point about being able to process only 4-or-so items at once, it seems you are on the money! I recently saw an old Veritasium video titled 'The Science of Thinking' which discusses this limitation. In the video he mentions how chunking enables you to parse/comprehend a larger total volume of information, but still in only so many total chunks at once.

    Timestamped link, if I may post it: https://youtu.be/UBVV8pch1dM?t=190

    1. Hi Michael,

      thank you for the comment.

      Yess!:) Exactly what Derek is talking about. I was basically trying to figure out what the prerequisite to chunking was and whether the concept of subitising was only related to telling the amount of something or it's something deeper.

  2. Great reading, Vladimir, I've even read it all loudly to fully enjoy it as being not native English speaker. Some comments to 'Possible relation between subitising and automation':
    I've read about an illiterate shepherd who was able to assess the quantity of the flock around 100 sheep in one glance. So this skill might be trained or is innate (and the shepherd was somewhere on the far side of Gaussian distribution) :)
    But generally, yes, you are right. I personally apply this concept to my language learning by splitting the unknown text into 'blocks (phrases)' of, say, 1 - 6 words depending on their length and my previous knowledge of them. Simply by writing vertical bar by pencil or pipeline character on keyboard. Then I use these building blocks for additional activities.
    And one more personal note to you: All your random thoughts are true pearls and you might better publish them separately or in smaller groups to get more attention and feedback :))

    1. Hi Jan,

      thank you very much for the nice comment:)

      About the shepherd, who was able to tell the quantity of his sheep, I suspect there something else at play here.

      Here is for instance what Wikipedia says:

      "Sets larger than about four items cannot be subitized unless the items appear in a pattern that the person is familiar with (such as the six dots on one face of a die). Large, familiar sets might be counted one-by-one (or the person might calculate the number through a rapid calculation if they can mentally group the elements into a few small sets). A person could also estimate the number of a large set—a skill similar to, but different from, subitizing."

      Thank you for the supporting words about my writing:) I will think about what you said. I do however enjoy the style Tim Urban has, (the Wait but why blog) - that is long posts once in a while, but shorter, more frequent posts have their place in the world too:)

      All the best.

  3. Great reading, Vladimir, thanks much!
    Just my personal note: your thoughts are true pearls, you might consider publishing them one by one or in smaller groups to get more attention and feedback :)

  4. Hello Vladimir, this was an outstanding and inspirational article. After reading it I started to search for more sources of subitizing :) Because I think it is great example, I would like to share with you one phenomenon, which I have experienced with learning russian. When I started learning russian alphabet, I noticecd, that I was able to read 3-4 letter words easily, but words with 5 or more letters were MUCH harder for me to not only to read, but also pronounce. After a while I could find somehow a way to improve at it, but only today my question was answered "scientifically". PS: I was not able to translate word subitise into slovak, do you think we can say "subitizovať"

    1. Hello and thanks for the nice words.

      Your experience with reading Russian words might be revolving around the same concept, but I really have no controlled, double-blinded, peer-reviewed scientific study to back me up on this:) It's just something I noticed over the years while trying to learn different things in different areas in a short amount of time and wanted to propose the idea for a healthy debate:)

      And interestingly, I experienced this 4-block limit when trying to improve my swimming technique too - meaning it wasn't something that would only be related to a purely mental task, but something related to muscle memory too. It seemed like, while swimming, I couldn't seem to concentrate on improving more than 4 things at the same time.

      In Polish, subitizing is 'subityzowanie', so maybe in Slovak it is 'subitizovanie' too, but I'm not sure.

  5. I really like how you mentioned the necessity of automation in functionality. I'm learning my fourth language (third foreign language) but with both my third language and this one, outputting the language especially in spoken form has been incredibly painful despite how much I'm able to understand. I figure a part of this is due to the lack of automation for both of these languages. When speaking we're juggling so many things at the same time (such as deciding which words to use, whether the set of words sound natural when they are together (collocations), how to arrange the words so the sentence is grammatically correct, how to pronounce each of the words etc). I'm thinking that at this point memorizing fixed expressions, collocations and whole sentences by native speakers may be more helpful for my output as opposed to individual words.

    I remember you mentioning that on a daily basis you're only actively maintaining your native language, English and Mandarin Chinese. I'm also curious about what you are doing for such maintenance. After not actively learning English for 10+ years I'm noticing that my active English has deteriorated noticeably, and so is my native language now that I'm living in the United States. Would greatly appreciate some input from you!

    Thanks again for another great post!

    1. Hello and thanks for the comment.

      "When speaking we're juggling so many things at the same time (such as deciding which words to use, whether the set of words sound natural when they are together (collocations), how to arrange the words so the sentence is grammatically correct, how to pronounce each of the words etc)"

      I came to literally the same conclusion when learning Mandarin. There are just too many things you need to think about in order to automate things effectively. For me, with Mandarin it was:

      1) pronunciation (sounds and tones)
      2) sentence structure
      3) which words to use/how this or that thing is said in Mandarin
      4) syntax
      5) grammar
      6) sentence accent
      7) my own accent
      8) the actual idea/information that I was trying to convey through Mandarin

      Apollo Robbins coined a term called 'attention currency':


      He basically says that you have a limited attention budget and every thing you consciously concentrate on 'costs' you attention currency. So by the time you're done with the above 8 points you are totally 'out of cash'. It's not just that you can't automate information chunks/blocks effectively, you can't even notice patterns in the language and reach "aha" moments because you are completely out of 'attention currency'.

      Memorising entire blocks of information is slightly better.

      "I'm also curious about what you are doing for such maintenance. "

      I just listen to audio in the target language every now and then and simultaneously translate it into my native Slovak language in my head.


  6. Very interesting. I read a paper about how the the early study of chinese character (as opposed to a 'postpone' approach) actually 'predicts' the future succes of the language learner. As a learner myself, I came to the conclusion that the most efficient language learning process is based on a multiple-sided-approach, starting from the good old four pillars (reading, speaking, writing, listening) all at once (does it count as 4 chunks?;). And talking about automation, when talking about language learning, I don't understand how it is influenced or related to 'understanding' nor 'subsiding'. Automation, I believe, it's pure repetition: repeat a task long enough and it will become 'automatic'. Understanding and, at a certain extent, subsiding are psychological needs that can influence positively the process by poviding psychological confort. That can help when writing, when we have the time to consciously think about the language, rather than 'automatically' produce it. It looks like the opposite of automation to me. I love the discussion anyway, and maybe I'll think differently tomorrow.

  7. Nagyon érdekes cikk! Sajnálom, hogy nem csinálsz több Youtube videót, bár teljesen megértem. Mi amúgy találkoztunk az utolsó pozsonyi Polyglot Gathering-en, amikor magyar téliszalámit kínáltam a magyar standon és te arra jártál megkóstolni. Remélem jól alakulnak nálad a dolgok, minden jót kívánok! Üdv. Berlinböl

    1. Szia Katalin. Köszönöm a kedves komentet. Lehet, hogy a jövőben készítek majd még néhány youtube videót :) Leszel a budapesti poliglott konferencián?

    2. Sajnos nem leszek ott, pedig biztos jó lesz. Nem nagyon járok már poliglóta találkozókra, leköt a munka, a család... De sose tudni, talán majd a gyerekek mennek egyszer ha nagyobbak lesznek, ök 3 nyelvvel nönek fel. Úgyhogy érkezik a következö poliglóta generáció! :) Jó konferenciázást kivánok Budapesten ha mész, és talán egyszer még összefutunk a poliglóta univerzumban! :)

    3. Koszonom szepen:) Sok sikert mindenben.