To learn a language

Created sometime in Dec 2023. Updated 06 Jan 2024.

I came across a method of language learning called AJATT (All Japanese All The Time) at the start of high school. I was a precocious just-turned-teenager who wanted to aggressively optimize every endeavor I took on (an obsessive min-maxer, in other words), and if it meant diverging from the conventional, then that was all the better. Of course, this was a precursor to the countless other times I ended up spending more time researching about doing something rather than actually doing the thing… but I digress.

Said method was espoused by an eponymous blog, which was and is janky, and which required registering an account to have meaningful access to the posts. It’s as ungainly today as it was five years ago, but the AJATT method lives on notoriously in the online language learning community. The author of the blog (Khatzumoto) didn’t invent its main tenets— back in 2001, a pair of Poles, Tom Szynalski and Michal Wojcik, laid out a similar method of learning English at a blog called Antimoon— but it was revolutionary for me, as someone who bought into the usual methodology of classroom drills, grammar rules, and vocab lists.

As hinted at by the name, AJATT was based around immense amounts of language input: listening to and reading anything you could get your hands on, whether it was anime, live-action shows, books, internet articles, radio, etc. But the most spartan aspect of the method was its insistence on cutting out everything not target-language related in your life. Though impractical, Khatzumoto challenged you to live like you were a native speaker of that language, with hours of actively and passively engaging with the language daily. On the scale of quality versus quantity, quantity eventually wins almost every time.

This mass-input approach goes hand-in-hand with conclusions drawn from research conducted by linguist Stephen Krashen. In his 1981 paper “Second Language Acquisition and Learning”, Krashen presents what he calls the “Monitor Model” for second language acquisition by adults:

...adults have two independent systems for developing ability in second languages, subconscious language acquisition and conscious language learning […] subconscious acquisition appears to be far more important. (italics original)

The subconscious system doesn’t require knowledge of explicit rules. It judges correctness based on “feel”. A native-level English speaker does not need to know about the intricacies of a case system to know that “Him loves she” is a grammatically impossible construction. On the other hand, the conscious system supposedly benefits from error correction based on a set of rules.

What Krashen calls the “formal knowledge” of the second language can inform the acquired system, but the user must be careful that his learned knowledge does not stifle, or worse, completely displace his acquisition.

By now it’s clear that traditional methods of teaching languages in classrooms are antiquated, depending far too heavily on language learning rather than acquisition. Therefore, the key skills needed to speak a language fluently, such as pattern-matching, that lend themselves to the necessary “feel” for proper grammar, can’t help but atrophy. This inevitably leads to the unfortunate phenomenon where a student has learned a language for many years (and indeed, in Krashen’s terminology, he has “learned” it!) and is able to, for example, decline an adjective so that it matches its noun, but still cannot speak the language with any competency.

But if the emperor’s been stripped of his clothes, and has been so since the 1980s, why hasn’t anything changed in classrooms? Why are we still so worried over outputting when we haven’t even gotten a significant amount of input under our belts, valuable input without which we can’t build vital schema deep in our subconscious?

One reason is that language acquisition is an apocryphal and highly individual process. How the mass-input approach should be implemented in classrooms is unclear, unlike the current reigning approach which has hundreds of years of tradition backing it. The huge amount of inertia aside, the mass-input-approach defers obvious benefits for the long-term. The student that spends an hour of class time listening to a podcast in the target language is not going to yield as tangible progress as the student drilling a vocabulary list (at least you can quantify the latter with the number of words recognized/learned). But iterate this process over months, and it will be the first student who has started to grasp the language with the intuition that is critical for fluency, while the second likely never will. It’s easy to see how this could put a damper on motivation, because a large part of input is learning to be comfortable with the unknown and incomprehensible— and we, as adult humans, desperately want to understand as much as possible, as fast as possible.

AJATT still plays a large part in popularizing the mass-input approach today, but it has also inspired a group of language enthusiasts and programmers who have founded the Refold method. Other big players include LingQ, a service made and backed by polyglot Steve Kaufmann, and the YouTuber Matt VS Japan (who I believe is one of the founders of Refold). I am always looking forward to what the people at Refold are building and am using their roadmap to teach myself Russian, in fact. If you are interested in teaching yourself a language, I highly recommend reading through their roadmap, and run with it without thinking about it too much.

Yes, I said that right: start without thinking about it too much. Language learning bears enough inherent inertia already. Some endeavors depend very heavily on starting conditions, forcing you to meticulously chart your course before you even take a first step. Language learning, in contrast, is far more forgiving.

Effective language learning is based on a very small subset of habits, so simple that you might have trouble believing that this is all there is to it:

Talent might play a small role in it (often, it manifests in people having a good ear for pronuncation/accents and therefore being able to reproduce native-like speech much more quickly) and so might intellect, whatever that means, but like I said before, it's a numbers game. Persistence is what is really at the core of success here; it isn't apps, it isn't expensive courses, and it certainly isn't ChatGPT.

Of course, whatever method Refold or whoever else (including me) purports is not a silver bullet, and absolutely none will get you to guaranteed fluency in X months. In fact, you can probably discredit the creator of any method that claims to do so as a seller of snake-oil. What I've hapharzardly laid out here is more of a philosophy than a regime, but it sets you up for life, and you can apply it to any foreign language, living or dead or constructed.

In the end, if you want to learn a language, you must listen to it as much as possible. Then, once you have acquired enough vocabulary, you must read as much as possible. Now iterate this over thousands of hours. How unglamorous and bald-faced is that? It might be difficult to believe, but that is what's really at the heart of things. Like in many other skills, perfect is the deadliest enemy of good.

⇤ go back to index