For decades, one work stood out as the dominant resource for learning kanji: James Heisig’s Remembering the Kanji. This book, originally written when personal computers were still a novelty and spaced repetion only existed in the realm of physical flashcards, innovated by introducing a system of mnemonics to aid in remembering kanji. The basic idea of this system is to decompose kanji into constituent parts, each associated with a particular idea such as “bird”, and use their composition in kanji to create stories that help in remembering the kanji. The prescribed program is thus:
- First, use volume 1 to remember the writing and “meaning” of 2200 kanji. It’s worth pointing out that kanji do not have anything that could be called a canonical meaning. Rather, RTK associates each kanji with an english keyword that give a rough approximation of what a kanji normally expresses. For the first few hundred kanji, Heisig provides the necessary mnemonics, but as the book continues these become rare and Heisig leaves the task of coming up with stories up to the reader. By the end of the volume, most kanji are introduced with no information beyond the glyph, its number, a keyword, and the keywords of its components.
- Then, use volume 2 to learn the readings of the kanji in volume 1. That’s right, as you trudge though hundreds of kanji in volume 1, you are not expected to learn their readings, and this is by design, with Heisig believing that learning writing first is more “efficient” than learning multiple things at once.
- Finally, use volume 3 to repeat the steps in 1 and 2 for 800 more kanji. Once again, the glyphs and their “meaning” are separated in part 1 of the volume from their readings in part 2.
The core idea, that of decomposing kanji and using mnemonics to commit them to memory, is fundamentally good. However, the same cannot be said about the methodology advanced by Heisig.
First, we must dispell the notion that learning less is more efficient. Language is, at its core, defined by relationships. When we hear the word “apple”, associations immediately come to mind: “apple” is a subset of “fruit”, so it does a thing called “growing” on things called “trees”. It is also a subset of “food”, so a thing called “eating” is often done to it. In fact, these associations are so strong that often you can guess a word without having any direct evidence for it: “Bobby told his teacher his dog ate his _”.
The same is true with kanji. If you show a native speaker 金, they will not simple think of the platonic ideal of metal, but rather immediately think of associations. Probably, first they will think of 金曜日 - friday, then maybe お金 - money, and then, why not, 金属 - metal. Forming these associations is at the core of learning kanji. Each association reinforces your knowledge of the kanji, and even when faced with an unknown word that contains known kanji, thanks to these association you can more often than not hazard a guess at the meaning and reading. Their individual “meaning” is, at best, a crutch for learning that will have little purpose when actually using the language.
Or consider this hypothetical. Imagine one student learns:
- 音 - sound
Now imagine another student learns:
- 音 - sound
- 音（おと） sound
- 発音（はつおん） pronunciation
- 足音（はしおと） sound of footsteps
The second student may need a greater initial investment to commit the additional information, but consider how much more reinforced their understanding of the kanji is now: rather than a single, flimsy, association between an english word and a glyph, there’s now a complex web of associations between three different words (in the target langauge!), two different pronunciations, and one english keyword.
Weaving such a web of knowledge is precisely the goal set up by Kodansha’s Kanji Learner’s Course. This often out-of-print book is a single volume work that teaches 2300 kanji. Each kanji is presented with one or two English keywoards, a mnemonic tool (yes, for all characters, not just the first few hundred), and some example vocabulary. Critically, the course is structured such that the vocabulary taught with each kanji only uses kanji that has already been covered. This means that not only does each entry weave a web on its own, but it even plugs into the student’s existing knowledge, further reinforcing their already-covered kanji.
Another alternative is the ever-popular website WaniKani, which fundamentally uses the same methodology as KKLC. The important differences is that WK is a website and it delivers its material in flashcard form. Aside from the monetary issue of it being a subscription service and thus costing more than KKLC if used in the long term, the primary issue I have with KKLC is its lack of flexibility: when I was learning with KKLC, I very often found myself creating mnemonics and keywords of my own that resonate better with me than what the course itself recommended (ironically, this is why RTK omits mnemonics altogether in later kanji). WK does not have this flexibility: you must learn its keywords for both kanji and components.
There is, to my knowledge, no real argument in favour of continuing to promote RTK in the wake of the advances brought by KKLC. The popularity RTK enjoys is, I am entirely convinced, entirely inertial from the position it enjoyed for decades as the dominant kanji learning resource on the market. It has since been rendered obsolete and should be gracefully retired from recommendation to japanese leaners.