“Don’t learn kanji, learn words!”

“you can’t ‘learn’ kanji lol”

Hang around in any Japanese-learning community for long enough and you’re guaranteed to see something along these lines. It’s advice that’s inspired by a long history of bad methods, and it’s given sincerely by people who genuinely want to help. It’s also something I happen to strongly disagree with.

Let me explain. I don’t think that the people who say these things, or the motivation behind it, are wrong. But I do think that “learn words, not kanji” is good advice, explained poorly, and that “you can’t learn kanji” is predicated on a bad definition of what it means to learn. I’ve tried to argue my case a number of times, and it never seems to go as well as I would like. This article is my attempt to put my thoughts in order.

If you don’t want to read the whole thing, you can skip down to the summary at the end. However, I would really appreciate if you did read the whole thing, especially if you think you disagree. This is a topic I feel strongly about, and I feel like the very premise itself is fundamentally misunderstood. I’ve put a lot of care into trying to present my case in a way that will make sense, even to people who might completely disagree at first.

I would like to implore anyone reading further to read with an open mind. Be cognizant of the fact that everyone thinks in a different way, and that none of us are wrong — until we try to force our own thought process on another person who is incompatible with it. I’m only writing this article because I’m constantly told my way of learning kanji is invalid by people who don’t seem to even understand it in the first place.

With that out of the way, let’s get into it.

Kanji have meanings

“Kanji don’t have meanings; words have meanings, and words are written with kanji.”

This is a fairly common sentiment, and it’s one I feel is easily shown to be incorrect. If you know anything about the evolution of Japanese, you’ll know that kanji were originally imported from China through trade and correspondence as far back as two thousand years ago. The introduction of kanji into written Japanese was a long and complicated process, but in the end, they’re used in two ways in the modern day; as ideographic indicators for native Japanese words, and as components of compound words that use an approximation of their Chinese pronunciations¹ It’s a lot more complicated than that, but let’s try and keep this simple..

While Japanese and Chinese have many areas where they differ (as one example, in Japanese, 手紙 means “handwritten letter”, but in Chinese it means “toilet paper”) they still have quite a lot in common when it comes to kanji usage, even thousands of years later. 月 means “moon” in both languages; 字 means “letter”, 危 means “danger”. And here’s the crucial point:

Native Japanese words written with kanji are written with those kanji due to the kanji’s meanings.

When the first person to try writing the word あぶない was trying to decide on a kanji for it, he didn’t make up a new character from scratch. He also didn’t pick a kanji randomly and then say “from this day forth, this character is how we write あぶない”. He picked a Chinese character which had something to do with danger, and then applied it to a Japanese word which also had something to do with danger.

Of course, there are lots of characters which have multiple meanings, like 本 or 代. There are also characters which are used strictly for their phonetic value, like 亜 and 伊. Not every kanji is easily defined by one “keyword”. But that doesn’t mean that no kanji can be represented this way. As we’ll get into later, kanji meanings are a key part of the written language, and it’s simply not the case that vocabulary words are some atomic element which can’t be subdivided.

Kanji are not words

“Kanji don’t have English meanings; the meaning of a kanji is its 訓読み.”

Similar to the previous point, this is another perspective which only tells part of the story.

In order to be precise about how ideographic² Japanese kanji are not purely ideographic. Let’s not get too deep into the weeds. characters work, we (unfortunately) have to invoke semiotics. 飛, for example, is not the word とぶ, nor the English word “jump”. It is a sign that represents a concept that is fundamentally beyond language; the action of traveling through the air without touching the ground. A Japanese native can say that 飛 means とぶ, and an English native can say it means “jump”, and neither one is wrong.

There are many words which can be written with one single kanji, and read with their 訓読み. 鏡, 水, and 心 are three good examples. However, there are also kanji which, while they do have a 訓読み reading, are not most readily associated with that meaning. For example, the 訓読み of 果 is は.て, but the original Chinese meaning is “fruit”, and the most common use of that character is in fruit-related words like 果物 or 果樹. Is the “meaning” of this kanji actually はて (extremity)? I would suspect most people would describe it as 果汁のか or 果実のか, despite the fact that か is an 音読み reading.

Furthermore, not every word even has a 訓読み reading. For example, 信 (belief) only has an 音読み reading; the word 信じる uses an old form of the word する to create a verb. There is no native Japanese word that carries this nuance; the closest you can get might be よすがとする, but this word is never written with 信. There exist kanji in Japanese which have no native meaning at all; these characters simply cannot be said to “mean their 訓読み”.

The kanji 笑 is not laughter. Laughter, itself, is not the word “laughter”. Words and kanji are both signs that signify things that cannot be otherwise expressed. The meaning of 笑 is わらう, and it is also “laughter”. Whether you name it after the Japanese word or the English word for the concept it represents makes no difference.

You can’t learn every word

“As long as I can recognize words when I see them, why should I need to be able to recognize kanji out of context?”

I think it should go without saying that there are many, many words out there. Learning vocabulary is a lifelong endeavor, and even if you do learn a common meaning for every kanji, that doesn’t necessarily mean that you’ll be able to automatically understand compound words just based on the sum of their parts. Some words use Chinese grammar, like 所謂; others have non-obvious etymology, like 言葉; still others use kanji which only ever appear in a single word, like 躊躇ちゅうちょ.

Still, it can be incredibly helpful to understand common meanings behind individual characters. There are thousands of words which follow simple composition patterns to produce meaning in a predictable way.

I suspect it’s uncontroversial to say that 水族館 is formed by first combining 水 (water) with 族 (family) to produce “marine life”, and then adding 館 (building for…) to mean “aquarium”. You could make the argument that 族 and 館 are both suffixes and therefore “words” of their own, but what about something like 連続? 連 is not a productive prefix, yet all of the words it appears in have something to do with “linkage” or “union”. Hopefully, the meaning of a word like 連死 or 連苦 should be obvious without even looking them up³ Don’t try; I made them up myself (something like “series of (related) deaths” and “one trouble after another” respectively).

However, it’s in fiction where this becomes especially relevant.

Fantasy and science fiction stories are full of made-up terms which will never appear in a dictionary. As just a few examples, you have 陽華突⁴鬼滅の刃, 蒼月草⁵葬送のフリーレン, 氷弓⁶黒猫と魔女の教室, 魔響教団⁷魔女と野獣… the list goes on and on. Each of these words was created with the goal of telling a story or describing the thing it indicates. Some of these words will be easily separated into smaller component words (e.g. 人体発火現象⁸炎炎ノ消防隊 is made of three words that can easily be found in the dictionary), but others will appear to contain full words which must be interpreted as individual kanji in order to glean the intended meaning (e.g. 血族⁹Bloodborne may appear to mean “blood relative” but actually means “tribe of blood”, and refers to a faction of blood-worshipping vampires who are not, in fact, related to each other).

I’ve talked to people who go so far as to say that fictional kanji compounds aren’t intended to “mean” anything, and are just supposed to look cool or sound unique. There are certainly cases where this is the case. For example, the name 烈怒頼雄斗¹⁰僕のヒーローアカデミア may include the characters for “hero”, “rage” and so on, but the primary goal is to simply spell out レッドライオット in kanji readings, and 斗 is no doubt used because it’s “cool” due to its association with 北斗の拳.

However, in my experience this kind of “meaningless” name is a vanishingly small minority. I think it’s much more likely that the people who hold this view simply don’t realize the hidden meaning and conclude that there isn’t one in the first place. If you make a habit of studying kanji meanings, and develop the ability to recognize them outside of words you can find in dictionaries, you’ll be rewarded with a deeper understanding of novel compounds both within fiction and without. If this isn’t important to you, it’s not a big deal — but when people make up these new words, they do have meaning, that meaning does not come out of nowhere, and those words are meant to be understood, and it makes no sense to act otherwise.

Partial knowledge is enough

Suppose someone were to ask you “how many people do you know?” You might think of friends, family, your partner, your coworkers…but when you start listing off names, the person interrupts you. “Sorry, that doesn’t count,” he says. “I’m not asking how many people you met, or whose names you know — I’m asking how many people you know deeply. Where you know at all times exactly how they feel; what they’re thinking; their most secret desires and shames. How many people do you know?”

Would any of us be able to say we know anyone?

This definition of “know” is perhaps useful in very specific contexts — the Delphic maxim “know thyself” comes to mind — but we can probably agree that most people would never use it this way. When I say I know someone, I’m not implying anything more than that their name exists somewhere in my memory. We might not even have met; sometimes I might even use “know” to refer to a celebrity, or even an author whose face I’ve never seen.

In Tae Kim’s article “You can’t learn kanji“, he makes a comparison to the phrase “I learned computer,” attempting to show how absurd it is to claim full understanding of a complicated system. However, I feel this is a disingenuous comparison. Indeed, it’s impossible¹¹ With the possible exception of kanken passers to learn everything there is to know about all kanji that have ever existed. But I feel that, once again, this is an overly broad definition of “know” that purposely excludes a more achievable goal.

When I say I’ve “learned” a kanji, all I’m implying is that:

• I can distinguish it from other similar characters
• I know what it means

By this definition, I “know” over two thousand kanji. Can I write them from memory? Can I list every word they appear in, every reading they can represent? No. But there’s no canonical definition of what it means to “know” a kanji. This is just mine.

What “knowing” and “learning” means will vary from person to person. Indeed, one person’s definition might be effectively impossible. But at the same time, another person’s definition might be easily achievable. It’s an inherently subjective argument and nobody can be objectively correct — but it is possible to be objectively incorrect; by acting as though only one definition exists and demanding that everyone play by the same rules.

Kanji vs Vocab is a false dichotomy

“Kanji are just a way to write words, so just learn words.”

This is, by far, the most common advice I see when beginner learners ask for help “learning kanji”, and I think it has some merit. It’s certainly borne of good intentions. But once again it misses some nuance, and can only be considered to be genuinely helpful when it comes with context for why people give this advice in the first place.

James Heisig’s once-ubiquitous Remembering the Kanji course is a common example of what not to do. Heisig infamously didn’t even know the word いぬ when he wrote his book, and his suggested method (which was once ubiquitous) was to learn all 3,000 or so characters with English names first, and only then to move on to learning their readings…which were also taught individually in the second volume of the series.

Methods like Heisig’s have fallen out of fashion in recent years, and I think that’s for the best. However, as is often the case when a bad method loses favor, an overcorrection has also occurred. It’s now commonly held that that you have to either learn kanji or vocabulary, and this is where things start to break down.

The idea is that learning words will also teach you kanji along the way, essentially for free, is a highly subjective point. Just speaking from personal experience, this method does not work for me. If I don’t pay attention to the composition of the words and characters, they don’t stick in my memory. Furthermore, I’ve met many people who thought that 働く and 動く, or 待つ and 持つ, or 幸い and 辛い, were written the same way and could only be distinguished with context — all while claiming confidently that you can learn kanji transparently just by learning words.

While learning dictionary words alone might be enough to take you a long way, it’s not a substitute for learning to recognize the characters when you see them in an unfamiliar context. Studying kanji individually doesn’t mean you have to be able to write them from memory — in many cases it just means knowing where to look on each character to disambiguate it from others of a similar shape.

Incidentally, while I often hear the argument made that “even native speakers don’t pay attention to kanji composition”, as far as I can tell this is a complete falsehood. Kanji puzzles in the Professor Layton series, games like Kanjile, the 秒で漢字暗記 series, or Daiso’s 毎日百均謎解き puzzle series are all ample evidence that not only do native Japanese conceptualize kanji as a collection of parts, but also that they can readily call that information to mind.

As I believe I’ve demonstrated in the sections above, there are many words you can only learn by knowing the kanji they’re written with. There are kanji which have atomic meaning but do not exist as standalone words. Studying vocabulary and kanji are both important. Neither is sufficient on its own.

Kanji in isolation, readings in context

In Kaname Naito’s video “You Don’t Have to Study Kanji“, he immediately starts the video off with a skit where he writes the kanji 生, then starts listing out its many readings before giving up in frustration. His point, which I agree with fully, is that even if you try to memorize every reading for a given kanji, you still won’t know which one to use in a given word — therefore, rather than studying kanji, you should simply study vocabulary.

Whether he intended it or not, the video feels like a strawman argument to me. The tacit premise is that there are only two possible options; you either study kanji and memorize all the readings (and still won’t be able to use them), or you study vocabulary and learn the readings automatically.

I want to be crystal clear. There is a third option that takes the best of both worlds.

Studying kanji individually doesn’t mean you have to be able to write them from memory — in many cases it just means knowing where to look on each character to disambiguate it from others of a similar shape. If you learn words by their overall shape, you might be able to read something like 食べ物 at a glance, but not recognize the same character in the word 主食. But if train yourself to pick up the subtleties of each character, you’ll find it much simpler to remember new vocabulary because all of the data points will relate to and reinforce each other.

I have never tried — nor have I even been tempted — to memorize a long list of readings for the kanji I learn. My kanji-learning method has always, from day one, been to learn the meaning of a kanji (in English), and at most one common 音読み reading. I learn vocabulary words in parallel with my kanji studies, and I learn the readings of those words as a whole. This is the method used by Wanikani and the Kodansha Kanji Learner’s Course. Ironically, KKLC is often presented as a superior RTK by people who would also say that you should “just learn words”. Yes, you should learn words. But the opposite of learning words is not learning kanji. It’s learning readings.

“Learn words, not kanji” is terse and to the point. It’s also bad advice. “Learn words, not readings” is much better.

In summary

• Kanji do have meanings, and those meanings are not always words
• Lots of words are created by combining kanji meanings, and knowing those kanji is the only way to understand them
• You don’t have to know everything about a kanji to say that you “know” it
• “Learn kanji” and “learn vocab” are not in opposition; learning vocabulary is good
• You should not memorize kanji readings individually

I don’t think I’ve got it all figured out. But I do know what works for me, and I know what doesn’t, and I know I’m not unique. It’s frustrating when people act like I don’t know what I’m talking about, or that I don’t understand their position. It’s especially tough when those same people demonstrate subpar kanji literacy while insisting that their methods are the only way.

Hopefully, even if we can’t agree, we can at least all understand each other.