Mnemonics : problems and solutions?

Exactly right. You’ve better expressed what I was trying to say: radicals are ultimately only useful as a memory aid. They help you remember the meaning or pronunciation of characters. WK forces you to prove you remember these aids (else they are truly pointless) but the specific word or idea they associate with each really doesn’t matter (and you can add your own associations on the first review).

It’s reasonable to mostly stick with WK’s associations for radicals, though: every subsequent mnemonic they give for kanji using that radical depends on that specific association. If you define your own, you’ll be forced to define your own mnemonic for all of the kanji that use that radical (not necessarily a bad thing).

IMO, that’s why WK forces you to review radicals (to the annoyance of so many on this forum). Some of the odder ones they provide are admittedly annoying (I can’t keep “black hole” and “death star” straight for some reason) but I learned to set aside my annoyance at missing review items long ago. I know that if I miss something enough times I’ll soon see it so often that I eventually recognize it instantly on sight, auto-magically. Trust the SRS!

In fact, I’ve developed a habit of typing “ge” if I can’t recall any item within five or ten seconds, radical, kanji, or vocabulary (why “ge” I can’t really say, but it’s never correct in either language!).

I think that’s an important lesson with any SRS: you must learn to shed your ego and quickly admit you don’t know an item. I suspect many “speedrunners” never experience the true power of this SRS. They get so hung up on “guru-ing” the new kanji at every level as quickly as possible that they feel they can’t afford a single mistake! By contrast, I’ve even intentionally entered “ge” even though I’m pretty sure I know the answer, simply because I think I need more reviews!

Re: your rules of thumb for 月:

There are several true linguists on this forum, and I’m in awe of their level of understanding and ability to express complex ideas. Using language to discuss incredibly nuanced details of a language tickles my appreciation of all things “meta.” But I lack the brain and the rigor to be an academic, so your “empirical statements” ring true to me. I suspect it’s like “rules” for transitive/intransitive pairs: there are so many exceptions it’s more like an occasionally accurate hint than a “rule”.

Lastly, regarding scripts:

I just installed the keisei userscript. It’s pretty awesome and I’m definitely keeping it, but I realized that the WK SRS auto-magically taught me many of these anyway. I already had しゃ、しょ、and ちょ in the same mental bucket, for instance, making recall easy, but と caught me by surprise:

One I really can’t live without is the 似合い visually similar kanji userscript, though.

The keisei script is great to help you guess the 音読み for related characters, but the niai script is fantastic when you keep missing a character and aren’t sure why. In my case it’s almost always because I’m confusing it with a visually similar character. Usually, both characters use a different radical, but sometimes my brain thinks two utterly unrelated characters are the same. This is where the script’s facility to let you add your own “visually similar” characters really shines. I’ve become utterly dependent on this script.

Anyway, I’m getting way off-topic, but this has been a fascinating thread.