Reflections on learning Japanese after two years of study

Back in July 2018, I developed an ambitious goal. I would learn Japanese by studying every day and then travel to Japan in 2020 to test my progress (the timing based mostly on work constraints). Now it is 2020, and my Japan trip has been postponed until who-knows-when due to COVID. Excepting a very small number of days where I did not study, I think I have more or less succeeded at the daily studying part. The fluency part, not so much. In fact, my command of the language is far less than I was hoping by this point. I still have essentially a topical understanding of Japanese speech found in radio, podcasts, anime, and the like. I can express my thoughts more or less if the listener has a lot of patience, but my spoken Japanese is that jarring think-speak-think-speak pattern of beginners. My vocabulary is disappointing. I cannot read without the constant assistance of Yomichan (incredible software, by the way).

Nonetheless, I retain a sense of continuous progress. Moreover, my progress during the most recent half year to a year has been greater than my progress during the initial year of study. At least, that is my feeling. It is a hard thing to quantify. I will not be quitting, if you are wondering about that by this point in the story. Now that I have a better sense of the difficulty, I am merely adjusting my expectations. I now suspect it will take me three to five years total to reach basic conversational fluency. And I still find it fun.


Previously, I attempted to learn Russian (college classes), Spanish (high school classes and self-study during surgical internship) and German (self-study during my masters). I even learned around 100 Kanji in high school which I promptly forgot. But I failed to develop any competency in any of those languages, and I can say only a handful of basic phrases now. Two years ago, I had a sort of motivational revelation. One, I am interested in linguistics and culture, and I obviously have been my entire life. Two, I am currently surrounded by a cohort of talented international coworkers who can each speak English fluently in addition to their native language and others. Consequently, I am a bit embarrassed to make it to my thirties monolingual, with multiple failed attempts at language learning under my belt. I do not want to add another failed attempt.

Additionally, a nagging but inspirational thought took hold. The reason I failed earlier was not methodological. Rather, it is anecdotally clear that just about anyone can learn a language. I believe the only reason people (myself included) fail is because they quit trying. If I had continued working on German bit by bit since 2010, now ten years later, I would probably have at minimum a very solid grasp on German. So, 2018 was the second best time to plant a tree (or I guess, fourth in my case).

Another complementary change in my thinking came about in the last few years that has also contributed to my motivation. I saw the phrase “mastery is a path” for the first time and it struck me. It answers the generic question, “When are you done?” with “Never.” Of course, I know I flit from one interest to the next. I know there is some plateau where my Japanese will fossilize, and I probably will quit studying. I am almost certain I will not be motivated to trick natives into thinking I was born in Kyushu by way of my accent, for instance. But I am not giving myself a time frame. I did not two years ago say I will do this in two years. I merely hoped I would due to some uninformed hubris.

Of course, I knew Japanese is quite linguistically remote to English. In fact, that is one of the major reasons I picked it. Conversational fluency in Spanish in two years is very reasonable. I have heard accounts of people who seem to do it in a summer. Setting this same expectation for Japanese is probably objectively unreasonable. The US Department of State classifies Japanese as a “super-hard language”, estimating 2200 classroom hours to reach “professional working proficiency” (whatever that is). This is in contrast to Spanish, which they estimate at 600 hours. Based on my experience with both languages so far, this is believable to me. By the way, their groupings of languages by difficulty is interesting to look at in its own right.

So I do not feel particularly bad about my progress so far. In fact, maybe what I can do is respectable for my current time investment, and enumerating my limitations at the outset of this post may have been too downbeat. I can read hiragana almost effortlessly. I recognize many common jukugo instantly (simple stuff like 説明). My vocabulary and pronunciation is good enough to express myself simplistically but comprehensibly to a patient listener. My ear is familiar enough with the phonemes, cadence, and common sinewy bits of speech to hear and repeat rapidly spoken phrases even if I do not know what they mean. All in all, not too shabby.

However, I do think my efforts have been and continue to be quite inefficient. Every couple of months I consider my current methods, scrap some, and try something new. Without rigorous formal training, I imagine this is somewhat unavoidable. I have had to learn to learn a language in addition to the Japanese itself. Based on the endless variety of learning resources and schemes out there, I take it this is par for the course. Though I cannot help but wonder, “What if I had been learning efficiently right from the start?” If nothing else, the answer offers a ready explanation for the various polyglots shilling their strategies—L3, L4, etc no doubt get progressively easier and quicker to learn.

What not to do

If you are interested in learning Japanese or some other language, I do not yet have a tincture to sell you. Once I succeed, I might follow up with some recommendations. Until then, I may be able to help you avoid the some of the mistakes I made by detailing my route so far. The remainder of this post is a rambling chronology of my strategies to date directed at the aspiring language learner. Learning resources I have used in no particular order include Pimsleur, JapanesePod101, language exchange apps like HelloTalk, Anki (including Core 2000, KanjiDamage), News Web Easy, and native non-educational Japanese content (anime on CrunchyRoll and Netflix, various podcasts, and internet radio). Conspicuously absent are classes, textbooks, one on one tutoring, and Japanese friends. I have spent at least one hour per day, often two, and on occasion four or five hours studying, not including any passive listening or casually watching TV with subtitles. I believe much of that time could have been better spent.

While there have been and no doubt continue to be many inefficiencies in my strategy, if I could change only one thing, it would be to strictly limit my usage of Anki to a small fraction of my daily study time no greater than 30% and ideally even less. Briefly, Anki is an appropriately praised, often recommended, versatile flash card application utilizing a spaced repetition algorithm based on an early SuperMemo algorithm. The problem is Anki is addictive. It feels like progress even when it is not because it allows you to replace a complex multifaceted skill with simplistic numeric targets. It has its place. But it takes time, and since there is a finite amount of study time in a given day, Anki time can easily leach from or entirely replace time you should be spending practicing listening or speaking. For me, that has been true, and I feel like I am just now correcting this error. Many times I was not really in the mood for Japanese but still spent an hour on Anki and called it a day.

When in medical school I developed a strong taste for Anki. Attend a lecture. Make flashcards. Memorize flashcards. Pass test. Repeat. Grades were good. I tried a handful of other strategies but they were less consistent and less time-efficient. So naturally, when I resolved to really buckle down and learn a language, I mathed it out a bit. X (hundreds) grammatical rules + Y (thousands) words = understanding. Understanding + a dash of practice (which could be put off) = fluency. Could not be simpler. Hopefully you are laughing.

So I went through Tae Kim’s grammar page by page, creating flashcards from the grammar points, example sentences, and vocabulary lists, memorizing by rote. It was laborious. It took months. It was mostly wasted time, as I will explain.

A major challenge to Japanese and Chinese is you cannot read phonetically like you can with many languages employing the Latin alphabet. You have to learn how 2000-5000 letters (Kanji) with a many-to-many relationship to set of phonemes combine in groups of 1-4 to form words. My approach was to just try to brute force memorize the appearance of whole words with flashcards. This is a terrible strategy, but it may not be immediately obvious why if you are good at memorizing. 先生 and 学校, for instance, are readily distinguished. But at this high level, I found my brain was memorizing general shapes, which is not good enough. Immediate problems arose with basic words like 待つ and 持つ which I could not keep straight. 開ける and 聞ける as well. Yes, those are different words. I realized for complex patterns like 試験, I was just memorizing the relative density of the strokes, and later on when I added a similarly complex Kanji like 鉛筆, they would conflate and cause me to forget both.

So I started skipping the vocabulary lists and instead worked in a prebuilt Kanji deck from KanjiDamage. I made it through Tae Kim ultimately, but I was failing my grammar cards over and over as well, and I had no context for of much of it. It was starting to feel unproductive. I had some vague idea of most of the fine grammar—such as, counters, append だけ to mean “only x”, 食べさせる is causative, etc. But I did not have a big picture framework. In Spanish if you know patterns like “Tengo que comer algo” or “Creo que…”, you are well on your way to forming useful sentences. How do you say “have to” in Japanese? How do you say “if”? I am still fuzzy on these basic patterns. Despite the answers being essential knowledge, these are actually very complicated questions with many grammatical patterns requiring a lot of situational awareness to use correctly. What I was doing with piecemeal isolated grammar study was functionally useless, and moreover I kept forgetting and re-memorizing the individual points.

I did this for months without listening or speaking. I began to feel I had invested too much time to be unable to speak at all. At this point, I recalled my experience with Pimsleur for German. Pimsleur is an audio course consisting entirely of repetitive listen and repeat drills that slowly build in complexity, broken into thirty minute lessons. It had me speaking a bit of German quite fast, so I gave it a go in Japanese. It proved very effective, at least initially. I went through one lesson a day, all five levels (30 x 5 = 150 days, somewhere around $500). At around 90 days into that course, I tried out a paid native online tutor, and we actually talked! It was a clumsy basic conversation, but I came away thinking that I can communicate if pushed. I felt I just need more words and expressions. This point was approximately one year in.

Pimsleur’s pacing is actually much too fast by the later units, and I ended up turning lesson vocabulary into Anki flashcards by repeated pausing and writing the phrases just to keep up. While it is explicitly and intentionally lacking grammatical explanation, I would check the Tae Kim explanation for points I was curious about. Between my positive experience with Pimsleur for German and Japanese and Learn Spanish with Paul Noble, a similar but superior course for Spanish, I currently believe starting day one with these kinds of audio only incremental shadowing drills is the best way to introduce yourself to a new language. Moreover, I wish that instead of rushing through Pimsleur (at its recommended, way-too-fast pace, mind you) nearly a year ago now, I had started shadowing from the outset, developed a strategy to create my own drills, and continued drilling every day for the entire two years.

Done with Pimsleur, but liking the audio format, I started up a few months later. Their short, focused lesson format and example dialogues are extremely good, and I felt that I was making good progress. The story was amusing. Somewhere into their intermediate lessons I realized I was forgetting most of it. I restarted it, going back through slowly and methodically like I had done with Tae Kim, making Anki cards from the written material. I am a one-trick pony after all. I am not sure if their grammar explanations better suited me than Tae Kim, or if I was finally developing, by shear daily exposure, a holistic mental picture of the language (just nominalize everything), somewhere around sixteen months deep.

In the most recent six months, after a seven hour road trip to visit my family listening to Japanese conversation podcasts I did not understand at all, I resolved to passively listen all the time, thinking this is how a baby learns—and I am presumably smarter than a baby, so why not me? (Spoiler: babies cheat because they have mommy to help them and do not suffer negative transfer from their native language). I set up Japanese internet radio and native conversational podcasts to play during my waking hours in the background. I think this has been helpful, even essential, but nothing close to a panacea. This has not facilitated much understanding or built my vocabulary. However, it has helped my ear resolve individual words between the all the はs、のs、というs、そうs etc, which I think of as the sinewy bits of language. In English, I think this would correspond to learning to hear grammatical cruft like “the fact that”, “with”, “of”. Also some words just come up so often you are forced to look them up, and at that point they are already memorized (Why is everyone always responding with たしかに? [Definitely, …] Why does every news report contain 詳しくは? [For more information, …]).

Dominating all of this, and such a blur that I do not know when I first tried what, has been daily Anki. Rarely less than 40 minutes a day. At one point, deciding abruptly raw vocabulary was my main problem, I tried to work in Core 2000, a commonly cited shared, prebuilt Anki vocabulary deck (consisting of 2000 common words, as the name suggests). I rushed through it, knowing most already. Yet the new bits did not feel like they were converting to functional vocabulary. Somewhere along the way I had grown impatient with KanjiDamage and rushed through it at 50-100 new cards a day, ignoring reviews before stopping entirely, telling myself I would learn to read after I learned to speak. I gave up on all the Tae Kim and Pimsleur cards I had made as well, allowing them to automatically leech away (an Anki feature after you fail to recall the word multiple times), and I eventually suspended the remainder while I was working on JapanesePod101.

The Anki disillusionment began to set in. I began to think thoughts such as, “If I have spent such an absurd amount of time on Anki, why do I feel like what I know well has little to do with Anki at all?” “How can I have spent so much time on Anki and not have a functional command of thousands of more words than I know now?” and “What if I had spent those 40 minutes a day talking to someone in Japanese instead?” I deleted everything. That was one month ago.

What now

I now regard the primacy I afforded Anki in my daily practice (usually greater than 50% of my time and some days more than 70%) as one of my biggest mistakes. You will find this warning all over the internet, and I want to add my voice. I would like to more thoroughly discuss my reasoning, and suggest an appropriate place for Anki which I am experimenting with now, but that will be its own post sometime later. In short, you learn words and phrases by imprinting an entire memorable context. Translational, emotionless word mappings are only useful insofar as they facilitate this process. A short TEDx talk by Gabriel Wyner explains this much better than I can with a powerful anecdote.

I also strongly wish as well that I had given Kanji the respect they deserved. I still believe listening and speaking is more important, but can you even communicate in the modern era without text messages? I love reading, and will almost certainly spend more time of my life reading Japanese than speaking it. If I had been diligent, studying 1-3 Kanji per day for the last two years, I would have a very solid foundation rather than my scattered recognition after at least two attempts to rush them. I am trying a mostly visual mnemonic strategy at the moment.

As I mentioned above, I wish I had been practicing speaking in some form the entire time. To this end I feel it is well past time for me to force the issue and schedule regular one on one tutoring. I should also have allocated most of my time to consuming native content in some fashion. I was always hesitant to do this because I normally do not watch much TV and the visual context seems almost a requisite in the beginning. I have been forcing myself to watch shows for the last month, not haphazardly but with rigor, and I already feel my functional vocabulary increasing.

If you are looking to learn Japanese, I wont recommend any specific course of action to you. I am still working toward my version of optimal. Yours may be different. I definitely do not feel qualified to preach a methodology, but I hope you have taken away something useful from my misteps.