10 years, 8 countires, 6 languages: What I've learned about learning languages

Edit: I submitted the post three times, it being deleted each time, before I found that one of the links was shortened and Reddit doesn’t like that. I fixed it, quickly added the title back, resubmitted, then realized I’d made two major typos in my frustration. The title should read: 10 years, 8 languages, 6 countries.

Hi!

Five years ago I shared a document detailing how I learned Japanese, and while that was well received, it was also 66k words long. I've since learned that more is not better: saying more with less is better. And that's hard!

"Less is only more when you know what more is, and make a conscious decision to step back from that." — Jacob Collier

So, this time around, I decided to try to condense those 66k words + an addtional five years of learning into just a few thousand words. (Edit: 4k words. Will try to condense more later.)

So, here are the most important lessons I've learned:

Think in ideas, not words
"A mediocre workout done religiously outperforms a perfect workout never done"
Going abroad is a force multiplier: if you're not making progress at home, you won't magically begin improving just because you uproot your life
You'll overestimate how much you need to know to begin doing cool things in a language and underestimate the gap between that point and fluency
You will learn as well as you need to learn to do what your lifestyle demands of you, no better or worse; if you’re stuck, light a fire somewhere
Achieving fluency means you know one more language; you'll be the same person you are now, for better and worse, plus one language
"There's only two sorts of problems: 51/49 problems and 100/0 problems," and most things in life (and language learning) are 51/49 problems.
Some things are best learned with less hours and more days; other things are best learned with more hours and less days; a lot of learning boils down to figuring out which things are which
Your brain will figure a lot of shit out by itself, if you let it
Knowledge is a spectrum, not a binary; this is at the root of many (most?) early learning hurdles

And the rest of this post is a brief elaboration (~300 words) on each of those points. I mostly just want to bookmark how I currently feel about languages for reference by future me, but I hope that some of it can be interesting food for thought.

・

1. Ideas, not words

While ideas often transcend languages, words get jumbled between them.

To give a super simple example:

In English, we say My name is Sami.
In Spanish, they say I call myself Sami.
In Russian, they say They call me Sami.
In Mandarin, they say I call Sami.
In Japanese, they say Sami {to be}

To give another:

In English, rain is heavy
In Russian, rain is strong
In Mandarin, rain is big
In Japanese, rain is zaa-zaa

The underlying goal is the same, but the route taken to achieve that goal is different. Furthermore, the choice to take route A vs B is often entirely arbitrary: rain in a storm is big, heavy, strong, and sounds like zaa-zaa, but speakers of different languages have for whatever reason ended up preferring one of these descriptors to the near-exclusion of the others.

In other words:

Don’t directly translate it’s 2:00 AM and I really want to sleep.
Instead, focus on the underlying ideas: telling time, connecting ideas, communicating desires, and intensifying a statement.

The "next level" here is that different languages will put those underlying ideas in different orders and may omit/include different underlying ideas. You're not learning how to encode your native language into your TL so much as learning to process the world as speakers of your TL do.

Another take on this idea from the angle of pronunciation:

When we speak with a foreign accent, what we do is we take patterns that we know from our native languages… and then apply them to {another language}. We don’t do it consciously, that’s just what organically comes to us. But if the patterns of our native tongues are different than those of {the other language}, the result is that {our message} isn’t going to be clear.

Maybe you know how to construct the sentence, the words are accurate and you don’t make any grammar mistakes… but if you don’t distinguish the right words, if you don’t stress the right words and put emphasis on the words that are stressed, you become unclear. {Pronunciation is about} recognizing your speech patterns and listening to how native speakers speak, which helps you to understand how {a language} should be spoken.

— Hadar Shamesh on Melody, Stress, and Rhythm in English Intonation

2. Mediocore workouts

You can spend a lot of time optimizing your routine, but none of that matters if you don't actually do the routine. In fact, it's a net negative (Relevant XKCD) to optimize things unless you're already spending a certain amount of time on them.

Here's an excerpt from Peak: Secrets from the New Science of Expertise in which two dudes interviewed a bunch of experts to figure out how they "do" their thing differently than amateurs. IMO the book boils down to this screenshot and this insight:

Research has shown that, generally speaking, once a person reaches that level of "acceptable" performance and automaticity, the additional years of "practice" don't lead to improvement.

Combining these ideas, we find ourselves with a pretty straightforward roadmap to overcoming the beginner stage:

First, just focus on doing: do, consistently, until you reach a point where you stop making progress
Then, start optimizing whatever it is that you do

If you (like most people, myself included) find that you tend to fail at the "do consistently" part of things, especially early on when language is something that takes energy rather than gives it, here's how you deal with that:

Understand the Habit Loop
Create a Trigger-Action Plan (TAP):
Let the structure inherent to your daily life carry you to success

The definition of "small" is "so easy that you’d accomplish it even if aliens invaded and you came down with kidney stones."

The way you find "small enough" is to start anywhere, then to shrink your goal/routine whenever you fail to fulfill your TAP for a single day. Your TAP is small enough once you're fulfilling it every single day, without fail. Once you're in the habit of consistently carving a slice of your day out for your TL, it's relatively trivial to expand that slice as time and energy allows.

3. Going abroad

Before I moved to Akita, my university connected me with a senior alum who'd gone there ther year before. We chatted about a bunch of practical things, but she also imparted upon me The Dream: she went to Japan speaking only English, then came back conversationally fluent in Japanese.

Unfortunately, that didn't happen for me.

It’s somewhat impressive, in hindsight, but I managed to construct a nearly impermeable English bubble in the middle of Japan.

Not speaking Japanese upon arrival, I made (truly wonderful) English-speaking friends
Despite living in Japan, I spent my free time in English on YouTube, Reddit, and Facebook
I took an overload of transferrable credits that caused me to spend the rest of my free time in the library, studying, in English
I found part-time work teaching — you guessed it — English

For all the Japanese exposure I got, I might as well have stayed in Iowa and taken a Japanese class. (Sounds asinine, but having lived abroad for 10 years, I've met many more people with stories like mine than with stories like my senpai's.)

To be blunt, moving abroad provides no guarantees that you’ll learn a language. It merely guarantees that you’ll have opportunities to use your language.

The thing is, unless you're learning a language like Noongar, the internet has already blessed you with a lifetime of "immersion" opportunities.

Put negatively: If you don’t push yourself to “immerse” in French while at home, you probably won’t begin doing so in France, either.
Put positively: I learned infinitely more Japanese in Russia and Taiwan than I did in Akita and Okayama.

If you haven't yet reached a level in which you can mostly do whatever you want in your target language, the most important thing you can do is to find a way to spend time in your target language every day. Quantity, eventually, becomes quality.

4. Underestimation, overestimation

This one's a twofer:

Overestimation —

Not all words are equally valuable.

Language follows what's called the power-law distribution (this is my favorite blog post of all time). You've heard of the 80/20 rule, but what's cool is that you can iterate upon it: 4% of words are used 64% of the time, and ~.8% of words are used 51% of the time. (Actually, 135 English words make up ~50% of English texts. See this discussion and the top 100 wordsfrom a similar list.)

Or, to see this in action, there are ~1,800 unique Chinese characters in Harry Potter 1:

80 characters appear 100+ times
~800 characters appear 10 times or less
~250 appear only once in the entire book

If you didn’t know two of these super frequent words, looking it up would interrupt your reading nearly as much as not knowing all of the rare words. Conversely, not knowing a few of those rare isn’t going to cause you much trouble.

And this matters because time is limited.

It's really easy to fall into the trap of "just 1k more words till I finish this Anki deck..." or "there's a series of 4 textbooks!" but the issue is that premade resources contain somebody else's idea of what's important for you to learn in your TL, and those things may or may not align with what you actually need to learn to do the things you want to do in your TL.

If you focus on the things that are important to your specific niche of interest, you can quite quickly (relevant to total time needed for X level) bring a language from "this is a brick wall" to "I can do this, with patience and Google."

(I personally follow drawabox's 50% rule, which ensures you learn teh basics but don't spend longer on them than you need to.)

Underestimation —

The other side of the 80/20 rule is that ~50% of the language gets used only .8% of the time. This is unfortunate because information density tends to be inversely correlated with vocab frequency, which is to say that the "rare" word you don't know in a sentence is disproportionately likely to be the key word you need to know to understand the sentence. (See the discussion on p24 of this article.)

You can try this yourself : if you understand 80% of a text, you'll likely get 0% of its meaning. Text coverage =/= text comprehension.

Furthermore, a lot of language is domain specific, and there are a frightening amount of domains. This is to say that Nobokov's favorite word is mauve, or that authors and genres have quirks and conventions. Reading a lot of fantasy will improve your ablity to read fantasy, but that doesn't perfectly transfer to things that aren't fantasy. There will be a learning curve when you step over to financial articles in the newspaper or casual texts from friends.

The intermediate stage of learning is independence — the ability to do pretty much whatever you want, with a bit of support or preparation.

The advanced (nevermind bilingual) stage is so much more than that — it's the ability to effortlessly do many things you aren't interested in doing, and would likely prefer not to do.

Even having passed the JLPT N1, the gap in ability between myself and my wife (who works in Japanese) is massive. My wife similarly laments that she's nowhere near "fluent" in Japanese. Hell, I make a living because people pay me to write things for them, but sometimes I read things by Raymond Carver or Amy Alice Munro that make me feel like Tarzan.

Deliberation —

You're basically infinitely closer to being able to do any specific one thing in your target language than you are to mastering your target language.

So, you know, you don't need to be fluent to do cool things in your target language; on the contrary, you approach fluency by doing a lot of cool things in your TL. If you're willing to look things up and can tolerate not understanding everything perfectly, you can jump into your TL pretty early. By doing the things you love, you'll build the specific skills you need to better do those things.

5. Necessity and progress

Here are three seminal studies on the concept of “full immersion” that I found insightful:

(Multiple, 1994) — Julie moved to Cairo at 21 and went on to become a “native-like” speaker of Egyptian Arabic
(Schmidt, 1983) — Wes, a sociable Japanese immigrant to Hawaii, became very conversational despite the fact that his English remained “broken”
(Schumann, 1976) — Alberto, a 33 year-old Costa Rican immigrant to the US, never really ended up learning English

That’s an unsettlingly broad array of outcomes! You could move abroad and make incredible progress, make none at all, or plateau at “good enough for a foreigner”.

In 30 Language Teaching Methods, Scott Thornbury sums that variability up with two main observations that seem reasonable to me:

Those who succeed while abroad willingly integrate into the community and/or spend a lot of time actually using their new language
Those who moved beyond “good enough” had situations or helpers which pushed them toward ever-increasing accuracy and competency

A big part of the intermediate plateau comes down to the fact that reaching the intermediate stage entails becoming able to do basically whatever you want in your TL, with support, but achieving an advanced level of fluency involves becoming effortlessly able to do a lot of things you probably aren't interested in doing.

And this means two things:

Most people could realistically do everything they want to do in a language without ever reaching a C level of proficiency. If you can accept that, your life will be easier.
If you can't accept that, you need to find a new way to light a fire under your ass. Paul Nation, a linguist I adore, has many wonderful suggestions in his free e-book The Four Strands of Language Learning.

6. Language progress, personal progress

I tried to kill myself when I was 17.

That's beyond the scope of this post, though, so I'll instead share this quote from Miguel de Unamuno:

An obsession with traveling comes from fear, not love; he who travels often goes fleeing from every place he leaves, not searching for every place to which he arrives.

I had a lot of self-work I needed to do, but rather than do that work, I set myself up for a lot of suffering by convincing myself that the conditions for my happiness and wellbeing depended on certain virtually impossible conditions being met. In a way, by hyperfixating on those things, I was choosing to be miserable instead of addressing the things that would actually move the ball forward for me.

This in mind, know that life is the same play/theatre production most everywhere; the backdrop just looks a bit different.

Learning another language doesn't mean that you'll become a cool person, make tons of friends, be proud of yourself, find a better job, be happy, or whatever aspirations and hopes you may be pinning to it.

If you achieve fluency in another language, you will be exactly the same person you are now, for better and worse. You will simply be navigating those betters and worses in two languages, rather than one.

7. Two kinds of problems

In an interview about AI with Neil DeGrasse Tyson, one of the engineers at Google's "Moonshot Factory" said that there are only two types of problems in the world:

With 51/49 problems, being 51% right is good. So if you're playing the stock market, and you can accurately pick stocks that are going up 51% of the time, you're about to be really rich.

Then you have 100/0 problems — where 51% is not good enough, and even 99% is not good enough. If you're trying to shut down a nuclear reactor in an emergency, you really need the 100% answer.

And that resonated with me because, when I thought about it, it occurred to me that virtually all language learning problems are 51/49 problems.

Around the same time I heard that, I discovered David Goggins, a 297 pound (134.8 kg) man with a crippling fear of water who became a decorated Navy Seal and now runs ultramarathons competitively. In an interview with Chris Williamson, he made a rather pompous comment that I nevertheless found myself inspired by:

It’s easy to be great these days because so many people are weak.

I imagined that there were an infinity of universes — an infinity of me’s. The odds are that one of these me’s was born Japanese, did a degree in Japanese literature, and won the Akutagawa Literary Prize. This me isn’t going to beat that me in “Japanese writing ability” because I’m not willing to make Japanese the sole focus of my life.

Importantly, when I asked myself how I could become 1% better, I was somewhat surprised to find that I had tons of ideas. Aspiring to be perfect was paralyzing, but aspiring to “beat the multiverse’s next best version of myself” was actually quite fun.

If you treat 51/49 problems as 51/49 problems, you'll save a significant amount of time and effort that you can put toward your 100/0 problems.

8. Hours vs days

// Will probably replace this one, since it's come up in several of these sections already

I learned to confidently read the Hangeul in about 90 minutes, and I know that because I spent ~5 minutes per day going through Drops' hangeul course most days while waiting for lunch at work, and Drops limits free users to 5 minutes per 12 hours.

The good news is that this took only 90 minutes!

The bad news is that it took 90 minutes over the course of a month.

This is an important opportunity cost.

Some things you do need to sit down and hammer out, but many things can be acquired more passively over time. A big part of learning efficiently simply boils down to making better decisions about how you delegate your time and effort.

9. Effort vs exposure

Somewhere in SuperMemo's wiki, talking of spaced repetition, Piotr Woźniak drops this little gem:

…To maximize the scope of what you learn, you should set target recall to 0%; in other words, don’t use SRS at all, just consume content at random. At any target recall rate above 0%, you are trading away some scope in exchange for control over what you learn.

What happens when you use an SRS like Anki is that there is an algorithm that nudges you to review a piece of information when it determines that you are 90% likely to recall it correctly. (If you enable FSRS, you can change that number.)

What's really interesting is that, when it comes to memory, 83 is greater than 90! In a very out-of-my-ass fashion, this is the case because:

Some words don’t stick, even with a lot of effort
Most words stick with a certain amount of effort
Some words stick, even with very little effort

And words are not equally valuable:

Some things we need to be able to actively recall
Some things we need to be able to recognize
Some things it’s enough to recognize that it’s related to food, or a positive/negative descriptor, etc
Some things it’s enough to know that we encountered it before and deemed it irrelevant
Some things that we don’t know, we can judge by context that we do not need to know
Most of our unknown unknowns we will never bump into, given our usage of a given language

Using a lower recall % means you review less frequently, but this won't really reduce the practical value you get from most words. (Imagine learning carbuteror vs just some part of a car.) This means that you can cover more content for the same original amount of effort, and you ultimately end up learning more things than you would by focusing more intently on a smaller subset of content.

And this leads to two important ideas:

Instead of doing 50 reviews of 10 words that just don't stick, you could instead spend those 50 reps doing 10 new words. Of that body of words, some will stick easily and most will come with effort. Redistributing your "reps" from leeches to new words means that more of your effort goes to things that will stick: for the same amount of effort, you learn more words.
Some words, you want 100% recall of; other words, it's fine if you don't fully remember them. By not spending unnecessary effort on less valuable words, you can focus more of your effort on the stuff that's useful given your unique goals. By spending less time reviewing unimportant stuff, you gain more time you can spend actually using your language.

This is cool to understand because it means that you don't need to know your target language as well as your native language to do the specific things that are important to you well. You need to know a subset of the language well, and while that's still work, it's not an investment of 20 years.

It’s also cool in that, eventually, scope is probably going to become more important to you than control. There’s a natural “quitting” point flash cards. That point is probably in different places for different people.

(The challenge, of course, is judging which items are more/less important.)

10. Spectrums and binaries

Knowing the translation of the word is the most shallow relationship you can possibly have with it. I talked about that more in this old Reddit post, but the basic idea is that each word exists within a complex web of associative meanings:

Some words overwhelmingly appear in specific strutcures (think lieu)
Most words are associated with other things (when you think birthday, do you not also think of parties, friends, presents, etc?)
Most words tend to appear before and after certain other words (as we discused, rain is big in Mandarin but heavy in English)

And this presents a dilemma for the upcoming learner.

You may often feel frustrated with a seeming lack of progress — I know this word when I see it, but when I try to speak, I never remember it! — but we make a lot of progress before knowledge becomes visible. I loosely see words as moving through a funnel like this:

You don’t know a word
You see a translation and now know that Spanish has a word for this concept, too
You know the word exists, but you don’t recognize it when you see or hear it
You usually recognize it when you see it within your app/etc
You always remember the word when it appears in your app, but you struggle with it outside your app, without the specific context of the example sentence you’re used to seeing it in
You remember the “feel” of the word when you see/hear it — whether it has a positive or negative nuance, that it’s related to a certain topic, etc
You recognize the word when you see/hear it, but when you want to say it, it always gets stuck on the tip of your tongue
You don’t remember the whole word when speaking, but you remember that it’s # syllables long, or that it contains the ahh vowel, or that it’s related to some topic
You reliably remember the word, but still struggle to use it because you haven’t yet learned all the words it habitually appears next to, the phrases its part of, the sentence structures it tends to appear in, etc
You can reliably use the word as natives typically use it
You can be creative and “bend” the word for artistic purposes, puns and jokes, etc

For a similar reason, progress appears to slow down at the intermediate stage.

As a beginner, you learn ten words and, like magic, can express ten new ideas: hot, cold, big, small, near, far, cheap, expensive, good, bad
As an intermediate/advanced, you often are learning to express finer shades of an old idea: Warm, boiling, tepid, scorching, sweltering, sizzling, tropical, blistering, sultry, humid.
The expended effort is the same in both cases — 10 words is 10 words — but the intermediate 10 words grant you many fewer additional degrees of freedom than the beginner 10 words did, which is less of a dopamine spike

This mind, Anki should be supporting your immersion, not replacing it…. Perhaps unless you’re a total beginner and need to get an initial grasp on the language, however transient.

I think this knowledge makes it easier to wean off of Anki in favor of whatever you really want to do in your TL, but also makes more palatable to take your daily dose of Anki at whatever intensity you’ve deemed worthwhile.

・

Thanks for humoring me, lol.

0 comments

LingoHello Language Exchange

skool.com/lingohello-language-exchange-9962

Language Exchange with real-time practice with native speakers. Clubs, Lessons, online meetups, Games and much more!!!

Host a 🌱 Community Meeting 🌱

LingoHello - lessons, Clubs, La...

LingoHello Learn - Ai Chat, Qui...

Leaderboard (30-day)