It’s been 3-4 years since I started studying Japanese, and outside of friends I’ve not really talked about the process of learning a language - or at least my process. This post will talk a bit about what my progress is like, what I’ve found hard and what approaches I’ve taken to learning using apps and self-study.
Well how well can I speak? Generally, I can get through most retail/service worker interactions usually pretty smoothly. I can also make some level of small talk and talk about things unfamiliar to me with wildly varying levels of success. But if there’s a lot of speakers and it’s not just one-on-one the flow of conversation can often be too fast for me and I get lost. I’ve also studied more grammar than I can use off-the-cuff so I’m working on strengthening more of that so I can use and understand it more readily.
But before we can talk about learning a language it helps to have a basic grasp of what the challenges of that language are. Languages can be vastly different and seeing what’s familiar and unfamiliar acts as a means of gauging how hard that language is to learn. It’s probably not a surprise that Japanese is considered a very challenging language for English speakers to learn. Let’s enumerate a bit on why:
Why is Japanese Hard?
These won’t be the only reasons, just a large number of the things that pop into my head that I’ve spent a lot of time working on or struggling with. I’m sure other learners have found other things hard and maybe some of the details about other languages and how Japanese relates are a bit fuzzy but here’s a general gist.
The Writing System
Japanese has 3 writing systems, hiragana, katakana and kanji. Hiragana and katakana (known together as kana) are phonetic systems so in hiragana ま is ma and in katakana we write it like マ. Katakana is just used for loanwords meaning you’ll eventually encounter “katakana english” which is just a string of English words maps to Japanese phonology. An example would be “sutoresu” (su to re su) is “stress” and “taipu” (ta i pu) is “type”. These can be a bit vexing as you often need to sound them out, drop vowels and think of what sounds might be skipped to shorten it to be easier to say if it doesn’t immediately sound like an English word. Japanese does naturally loan from other languages as well with French, Chinese and Thai immediately coming to mind.
A lot of learners (me included) find hiragana easier to learn, katakana have more letters that look similar like シ (shi) ツ (tsu) ノ (no) ソ (so) ン (n) which can make learning hard.
For historical reasons は is normally “ha” except when used to denote a subject where it becomes “wa”. Another similar example is へ which is normally “he” but can become “e” if used as a particle.
Kanji is a whole other beast, based on the traditional Chinese writing system each character is a word and they can be joined to make words, or joined with katakana or hiragana. Japanese however comes with some unique challenges which come from simplification/modernisation of the writing system changing the kanji and not being a tonal language. This means the relationship between the kanji and how they’re said is weaker than Chinese and this removes some of the potential intuition and turns up the dial on “just memorize it bozo” territory.
Kanji also have multiple potential readings, there’s things called “uncommon readings” which there’s zero pattern for and you just have to memorize.
Why use kanji then if they’re so hard? When reading a language with no spaces you need them to disambiguate between words. If we look at this sentence with kanji 母は早い versus a version with just hiragana ははははやい we can already intuit some of the issues with removing kanji. Why not add spaces? My guess is it wouldn’t help in every case given the amount of homonyms in Japanese and it being a context based language. For example all of this kanji have a reading of sei 生, 性, 製, 製, 精, 聖, 姓. We’ve gotten rid of the pains of learning a tonal language, but at what cost?
Why does Japanese mix kanji and kana? Simply put Japanese has a lot of ways of conjugating verbs, joining words and particles which don’t work with kanji. These seem to not exist in Chinese hence why they can stick with kanji only.
Note above a kanji you might get the hiragana pronunciation this is referred to as furigana and used to assist readers.
Register
Japanese has different registers this is referred to as keigo (respectful speech) and you use different verb endings, different forms of words to communicate. Most apps will teach you polite form, there’s a more casual form also often taught and then more respectful ones which apps generally won’t teach but are more useful in professional contexts.
The general registers are:
- Casual
- Polite (Teineigo)
- Respectful (Sonkeigo)
- Humble (Kenjougo)
- Aggressive and vulgar (anime antagonist)
For introducing myself we can do this as:
- Casual: ダニエルですよろしく
- Polite: はじめまして。ダニエルです。よろしくお願いします
- Humble: はじめまして。私はダニエルと申します。どうぞよろしくお願いいたします。
Respectful doesn’t make sense with introducing yourself (it’s more related to the status of the person talking to you). But we can generally see how the more polite you get the more words you use. Also we go from ダニエルです (lit: it’s Daniel) to 私はダニエルと申します (roughly: I am called Daniel). Lastly our “onegaishimasu” (お願いします) turns to onegaiitashimasu (お願いいたします).
You should generally know at least casual and polite. If you only know polite you’ll end up confused when in a more casual conversation - speaking from experience. Obviously, if you want to work in Japan you need to go deeper.
Counting
Japanese when counting things has “counting words”. This means that the word for one (ichi) will change based on what you’re counting! For a short example one will change as follows when counting the following things:
- People - hitori
- General purpose objects - hitosu
- Small animals - ippiki
- Rabbits and birds - ichiwa
- Flat sheets (paper etc) - ichimai
- Bottles - ippon
Counting words are numerous, generally I fall back to the general purpose object counter for counting things that aren’t animals if I don’t remember. But it adds more complexity to the language.
Actually, a current evolution of the language seems to be younger people are counting rabbits using the small animal counter. So we might see in future years rabbits being counted the same as birds disappear from the language.
Pitch Accent
When speaking, pitch accent or intonation is the change of pitch during a word. In English we use stress-accent, so the stresses in words impact timing and help us distinguish words. In Japanese you have a word-level pitch change. This contrasts to Chinese where it’s more like a character level pitch change. It can also help to distinguish between words with the same pronunciation.
The Youtuber Dogen has a lot of sketches around this and a Japanese course (and maybe an app now) which teach this primarily. As far as I’m aware none of the major language learning apps even address this. But it’s an important part of getting things sounding right when speaking.
How I Learn?
I started off on apps gradually adding more, anki and immersion based learning. I’ll go through all of these roughly in order with some feedback/thoughts on the pros and cons of each one.
Duolingo
The elephant owl in the room. Duolingo has been written about by a lot of people as a bad
way to learn a language and that’s ultimately true if it’s your only tool. The issues ultimately
boil down as follows:
- Poor quality TTS
- Lack of variation of grammar and form
- Multiple choice can be too easy for pattern matching brain
- Feedback is limited and often poor quality
With all this what does Duolingo do right?
- Vocabulary spam
Out of other apps in a similar space Duolingo hits you with way more nouns, verbs and adjectives than the others. You can spend a month on the same repetitive grammatical pattern but get new words every week. This is largely the only benefit I find, vocab bootstrapping.
Now for a breakdown on the negatives.
Poor Quality TTS
Duolingo suffers from the Google Translate failings in Japanese with it’s TTS. Often it gets the pronunciation of a kanji incorrect or sometimes mispronounces は in a word misinterpreting it as the subject particle. A recent example I got was matching kanji to pronunciation where it gave me 体 (karada - body). In words like Physical Education this kanji is pronounced tai but on it’s own it’s karada. The kana pronunciation hint said karada the audio said tai. This sort of mistake is common in duolingo. The furigana hints are always correct but the TTS doesn’t always match - Google translate does the same thing as well.
Lack of variation of grammar
Japanese has so many verb conjugations, and there’s so many ways to form compound sentences. Duolingo explores very little of these and mainly only in polite form. I think it took about 3 years to get a lesson that used the casual form and it doesn’t consistently teach it. So for new verbs I don’t always get what verb form it is just how it looks in teineigo form. For any word you learn in Duolingo you should probably look it up in a dictionary and explore some sentences with other forms if you want to actually retain it.
For an example I used to really struggle with casual form. What Duolingo would give me for “can you do this?” would be something like:
これができますか？ Kore ga dekimasu ka?
When I was on the farm with my brother in law he asked me like:
できる？ dekiru?
Also things like “is it tasty?” you can do more polite (Duolingo preferred)
美味しいですか？ oishii desu ka?
But my wife has used with me:
美味しいのか？ oishii no ka?
I’m on a >1200 day streak on Duolingo and it’s never given “no ka?” as a way of turning something into a question. This isn’t advanced grammar it’s just casual and it should appear much much early in Japanese study.
Pattern Matching Brain
If you do Duolingo long enough you can pick up on the patterns they use, “pick all the words that meet
Bad Feedback
When I started Duolingo didn’t have any speaking exercises for Japanese. It now does but the speech recognition is so bad on short utterances they’re ultimately a waste of time. There’s no feedback on pitch-accent.
Busuu
Busuu is another app like Duolingo but it has a few benefits:
- Videos of people saying the sentences as well as TTS
- Better grammar explanation and more variety of grammar
- Casual and polite forms
- More words related to Japanese culture and festivals
- Native speakers can grade your writing or speaking
This is a higher effort to create course, but I did complete it and despite what they say I don’t think this can be considered B2. There’s just not enough variety in topics and vocab. But it does try to teach grammar which I prefer.
Naturally it has some negatives:
- Much smaller vocab and sentence variety
- The native speakers are other users so you often get no reviews. Reviews you get are short like “perfect” if you don’t make a grammatical mistake
There’s not really much more to say on Busuu. It solves some issues with Duolingo but there’s a lot less content. I think it’s a good companion to it and I think I completed the whole course in roughly 450 days.
Kanji Study
I use this app to study Kanji Japanese Kanji Study. It’s spaced repetition with sentences so fairly normal. I’ve turned off the writing tests as they’re a bit ropey and I don’t expect to write by hand much.
Kanji are good at unlocking words meaning without knowing the words. And knowing some helps acquiring new words when you encounter them because you can grasp at the etymology even if vague. Because of this I study kanji separately with this app and don’t necessarily focus on readings but meanings.
There are some negatives though, I’m not sure how they’ve picked the kanji ordering it seems to be from some curriculum or scraping. But the order is odd. Kanji like halberd and arrow show up before voice or eyebrows. I don’t know about the rest of you but I use the word voice more than halberd.
Also it has some weird words that are kinda archaic and no Japanese natives I’ve met know. One example is: 破瓜 this means 16 year old girl, 64 year old man and a girl losing her virginity. The rightmost kanji is melon, but 八 is 8 and it seems some poet from ancient china thought if you combine two 8s you can get melon so 8+8 or 8*8 make sense as a persons melon year I guess. It’s insane that this pops up before the kanji for “know” and most of the colours. In fact the only colour I saw before it was white. I’m studying 663 kanji on this app and I still haven’t seen the kanji for blue and I only saw green last month!
Rant over, this app is reasonably good at explanations and drilled repetition. Kanji choices are weird sometimes but most of them are somewhat normal. I’m in two minds on whether proper nouns and names should show up in example sentences as they become harder to intuit.
Anki
I’m using the JLAB deck’s which have grabbed content from anime, TV and movies and tries to introduce concepts following Tae Kim’s Japanese Grammar Guide.
I find anki is one of the highest reward ways of learning, but it has a higher time commitment and falling behind in the backlog makes it hard to maintain motivation. Also, self-grading yourself is hard and I sometimes have to look up further to see if my answer is acceptable because the back of the card lacks sufficient details.
But a plus is this is real voices not TTS and real sentences in the wild from media. And not just anime which helps as well!
Local Immersion
I have spent roughly a year cumulatively in Japan mostly in non-touristy places with not a lot of English (you may recall my rice farming if you’ve been here before). This is naturally not obtainable to everyone and different people might have different levels of comfort with travelling. I hear there are walking tours in Tokyo which are mixed language for some language practice.
No negatives, the only potential one is you or the native speakers being shy and not communicating to the full potential. But that wasn’t so much an issue for me. You’re learning a language you’re allowed to make mistakes and sound like an idiot now and again.
There are also meetup groups in most cities where you can learn the language with other learners. I haven’t really tried them so can’t comment too much, but I have been to some Japanese expat events in London with my wife (she’s Japanese) and they were good practice for listening and speaking.
Youtube Content
I need to do more of this for sure. But one channel I’ve watched a few of is いろいろな日本語 (all sorts of japanese). This channel uses comprehensive input and I’ve mainly watched the yotsuba-to videos.
The way this works is he’ll read the manga, and when reading it he’ll explain things about it and try and get you to understand only using Japanese. For example, volume one he’ll open it up and there might be a panel of yotsuba in a car being driven by her older brother. In Japanese you’ll get commentary like:
This is yotsuba, she is a child, she is a girl. She is in a car, this is her older brother. Older brother is boy. Yotsuba is younger sister. Outside is countryside, this is road. Yotsuba is small. Car is big.
There’s a lot of repetition, pointing at things, finding different words and simple words to describe them. It’s in my mind one of the best first steps to consuming video content entirely in Japanese and moving beyond English based learning.
The only thing that stops me doing more video stuff is the time I have to allocate to watch a video and actively lock-in. Often finding things that interest you is the best way to do this when you get more advanced. At the start you might just want to google things like “Japanese Listening Practice” and consume the equivalent of preschooler content.
But some examples of other content that I’ve added to my Japanese to-watch list to try and engage me:
- Miho nonaka professional climber doing a vlog
- Documentary on marriage equality in Asia
- Hiking in the rain
Journalling
I won’t be releasing this but I’ve been working on journalling in Japanese. My general process is:
- Write journal entry with minimal referencing other things - I will look up spellings if I know the word
- Review, look up things I wanted to say but couldn’t. Make sure I can understand myself
- Another review with AI assistance see how close/far I was
- Potentially rewrite parts to be better
I am still on my first journal entry for a trip I did to the North of Japan but I’ve covered about 5 days of the content. Writing takes so long and it feels so intense to be trying to grasp for what I want to say without a conversational partner to bounce things off and help me. I need to find more time for this but my last big chunk was 3 hours on a train when I could lock-in.
Here’s a bit of an entry feel free to translate it separately and see what I’m saying. I haven’t checked my work yet so it might be a bit rough!
韓国の後めいちゃんは宮古島に戻った。俺はバクパキングを始めした。
俺の計画
１．新潟に行く
２．お酒を飲む
３．山形に行く
４．即身仏を見る
でも青い森も行きたいですね。さいきん強いな地震がしたとたぶんだめと思いった！ 東京から新潟まだ新幹線に乗りった。チェクインまだ新潟のアニメ博物館に行った。 後で博物館、白山(はくさん)公園と神社に行った。神社でフィシングのおみくじをした。 しかし強いな風があったとフィシングは難しいでした。でもおみくじはいいです。 フィシングのおみくじは三分ぐらいかかりました。ずっとおじさんは俺を見た。“ 頑張りました”と言った！少し話していました。
AI Assisted Learning
One thing with the apps, is you often can’t combine data from multiple sources. But Busuu and Duolingo both have a list of words they’ve introduced to me, and grammar points. I wanted this data, so using codex I was able to export out those lists and build up a bit of analysis.
From the analysis I’ve encountered across all the apps I use for learning ~3600 words and have been exposed to around 1000 kanji. But more interesting, I asked it where there were gaps in my conversational Japanese and ended up with a list of connectives and some casual grammar points I’d not yet hit.
Using those points and the vocab I already had I got it to create an anki deck using a free Microsoft TTS for the voice. I did hit the Duolingo and Google Translate issue where it picked the wrong reading for a kanji in what feels like a very obvious way but I was able to fix it by replacing the kanji with the kana in the request.
I think uses like this and checking your understanding via back and forth with an LLM are incredibly valuable. However, LLMs often struggle with wordplay and humour particularly in other languages. So here’s an example where ChatGPT 5.1 wasn’t able to explain a joke in Naruto which a Tumbler post did!
me: There’s a naruto joke where a character is referred to as a nipple friend, but is corrected as “childhood friend” it seems to be based on a similar sound of nipple chikubi with childhood. What’s the word for childhood?
chatgpt: The word you’re looking for is 幼なじみ (osanajimi).
- 幼 (osana) = young/childhood
- なじみ (najimi) = familiarity / acquaintance So 幼なじみ means childhood friend.
The joke works because:
- 乳首 (chikubi) = nipple
- 幼なじみ (osanajimi) has a somewhat similar rhythm/sound pattern in casual speech, especially when characters mumble or mishear it So the gag is basically a mishearing or deliberate wordplay between something innocent (childhood friend) and something embarrassing (nipple), which is very typical of anime humor.
me: No it was something like chikuba
Chatgpt: <continues to tell me I’m wrong and it’s osanajimi>
Now there’s no way that chikubi sounds like osanajimi, in the end I gave it the kanji for chikuba after I’d looked it up and got the “oh right of course this is…” where it gave the correct explanation.
I just tried this with chatgpt 6.1 sol and it gave the correct answer after 3 internet searches. But with certain slang or wordplay things I have found LLMs to fall short as recently as 2025.
I also used AI to put together a Japanese grammar revision page, with examples and vocabulary from my study materials. The English prose has that LLM tone which is a bit grating however as quick grammar reference cards with some sentence examples from my various study tools it’s quite useful.
Conclusion
When learning a language a lot depends on what your aims are. My primary aim is to converse with native speakers but I’d also like to be able to understand Japanese media - manga, anime, film, music and more. I know people who don’t touch reading at all or speaking, just listening comprehension for anime. I guess if there was any advice it would be not to rely on just Duolingo, try to get into anki and then the more things you can add that are relevant to your aims the better.
Usually with languages the main issue is not trying or spending too long deciding the best way to learn. But I think if you can stay consistent and have enough variety eventually something useful will fall out even if it’s not the “most efficient” way.