Why does WK not provide the largest radical parts?

Last night I was reading a story that used 運転する which I learned a long time ago from a vocabulary perspective, but I’d never really taken the time to learn the kanji themselves yet. So I began digging into 運.

運 (Carry)

Why does WK not optimize for using the largest radicals that compose individual kanji, in this case, 軍 (Army)? Frankly, I feel the underlying WK data seems to create more confusion than it’s worth. Which makes more sense for internalizing the concept of carry?

  1. Scooter + Forehead + Car
  2. Road + Army

Option 2 makes considerably more sense to me…Armies carry stuff on the road


You learn “carry” at level 10. You learn “army” at level 15.

That suggests the order of their data is flawed additionally.


You can probably find some example that makes it easier, but what about the other hundred words connected to that radical? Don’t you think that would mess up the whole system? They can’t build a system around a couple of radicals and expect the rest of it to work out fine.

WK knows what they are doing for the most part, it is well thought through. They didn’t just wing it.


What word are you more likely to use in casual conversation, carry or army?

1 Like

These words aren’t mutually exclusive, army is a subcomponent of carry.

It would definitely have ripple effects. Everywhere that forehead + car was referenced should have been written with army from the onset for ease of retention.

1 Like

Yeah that is true. They probably have a good reason for it though, it’s pretty low hanging fruit.

It is also the case that in a system covering thousands of characters there’s likely to be some choices lurking in there that were just arbitrary, or which rest on somebody’s “well, to me keywords X+Y+Z seem like they make an easier mental image than X+A” opinion. As long as there aren’t too many idiosyncratic decisions it’s not a big deal.

I think I’ve mentioned before that Heisig gives himself a bit more leeway for some of these oddball choices in RTK by his writing style being more explicitly about what he found worked for him rather than a neutrally-phrased description of a system; and also by encouraging users to do their own mnemonic creation. Crowd-sourcing of users’ mnemonic stories also encourages sharing of “this split of components worked better for me” approaches (but comes with its own pitfalls too!).

1 Like

You also need to remember that WK is not responsible for the structure of the Japanese language. The “official” radical for 運 is “walk,” which WK has renamed as “scooter.”

And while “car” is a “true” Japanese radical, “army” is not.


I never said that WK was responsible for the structure of the Japanese language.

The official meaning of the radical for the kanji in question has multiple interpretations, road and walk are both equally correct. See this list. I’m well aware that WK decided to rename it scooter.

The term radical is used synonymously with subcomponent for the purposes of this site, as army is a building block of the greater kanji in question.

1 Like

You’re missing my point, which is not the name of the radical, but the choice of the primary radical.

WK is not defining building blocks in isolation. There is an entire structure of Japanese (and Chinese) linguistics, according to which 辶 is the primary radical for this particular character.

1 Like

With respect to Heisig, while I feel his system was particularly flawed (unique keywords are definitely challenging across the spectrum of the compllete set of Joyo), he did appear to agree with my recommendation for this particular scenario: 運 - carry - rtk-search


I don’t care what the primary radical is for the kanji, never said I did. I have only ever referenced the complete set of building blocks that WK is using versus what make more semantic sense and result in fewer building blocks to need to remember. It appears you are the one missing my point.

1 Like

Yes, it’s obvious that you don’t care. But WK apparently does, hence their choice of building blocks.


You appear to be more interested in picking an internet fight than addressing the topic of the thread. For reference, “don’t care” doesn’t always mean literally “don’t care”, it can also imply not relevant to the conversation.

1 Like

If WK cared about fidelity to the traditional radicals they wouldn’t have renamed this one as ‘scooter’… 100% fidelity to the traditional radicals is IMHO not a useful goal for a modern mnemonic system anyway – where they get in the way the system should favour consistency and memorability, and IME that’s exactly what they do.


It really comes down the philosophy behind the system.

WaniKani wants to introduce useful kanji early and keep the total number of radicals one has to learn to a minimum.

RTK on the other hand uses a building block approach. Radicals and kanji that are subsets of other kanji are learned first, and then combined to make more complex kanji.

For instance, with 瀬 (“rapids”) WK tells you it is “tsunami” plus “bundle” plus “geoduck”. RTK would call that “water” plus “trust” because if some subset of a kanji is a kanji in its own right, treat that contained kanji like a radical and keep the number of elements in the story for rapids down. “The water in the rapids can’t be trusted” is so much easier to remember than some story about tsunami, bundle, and geoduck.

Which way is better depends a lot on your learning style, as well as your goals. The RTK approach makes for less total work to get familiar with a large set of kanji, but does so by prioritizing kanji that will help you learn other kanji, whereas WK focuses more on prioritizing usefulness for reading. Heisig kind of assumes you want to pause your other Japanese learning to quickly familiarize yourself with the kanji, whereas Koichi wants you to learn kanji over a longer period of time, but learning vocab and doing reading during that time.


Interesting take on things Corodon. I personally was opting for a hybrid approach between both systems (RTK and Wanikani) since I do a lot of active reading of Japanese to hold my interest and make note of in-context vocabulary I come across and work backward from there to dynamically learn that particular set of kanji.

Given this approach, I’m not interested in the reading/vocabulary portions of WK. I was testing out using WK’s dataset as an a-la-carte for learning kanji meanings on-demand, and working backwards if it leverages unfamiliar subcomponents. Given the plethora of ways people use these systems, I think the conclusion I’ve drawn is I’m probably better off creating my own as I go (which isn’t as much work as it sounds like with tools like Jisho.org and mobile apps like Shirabe Jisho).


If that’s your goal, you’re definitely better off building your own. Whether you agree or disagree with it, WK definitely has their own progression. They don’t even attempt to support an a la carte approach.