@SaiyaJedi:
That's good for the most part, but the shin'nyou radical (the one in 道), while written with two strokes (辶) in simplified characters and with three strokes (辶) in unsimplified characters, is actually a seven-stroke radical (⾡) and is classified as such in typical kanji dictionaries.
By the same token, 阝 on the left side (kozato-hen) is classified under 8 strokes (⾩), while on the right (oozato-tsukuri), it's 7 strokes (⾢), despite being written with three strokes either way. Or how the "grass" radical (kusa-kanmuri), typically written with three or four strokes (艹、艹、艹), is classified under 6 (艸). …There are more than a few like that, actually.
Well, when I said that I was talking in terms of total character stroke count, which as far as I've ever seen in modern classification, does not take into account the strokes in traditional radical classifications. Wow, I almost confuse even myself with that sentence. Well let's take concrete cases.
In my 漢字源 you can search either under 辶 (3) or ⾡ (7) and get the same subset of kanji. However, when counting the characters' total strokes, both cases use 3 for the radical portion. In other words, 込 is a 5-stroke character even though it can be found under the 7-stroke radical ⾡. But nowhere considers it a 4-stroke character even though technically it is written in 4 strokes.
Now, I've only used a couple different traditional-style kanji dictionaries, but it has been my impression that modern ones either show both radical versions (or one redirects to other), or they just use the the same number of strokes to categorize the radical as they do when counting total strokes.
@DarthAsthma:
I solved this problem by simply learning compounds along side Kanji. When you see that 間 in words like 週間, 時間 or 期間 you kind of get that one of the on'yomi is かん. I've found this method to be the most efficient right now letting me learn quite a lot of new words and even allowing me to guess often readings of new compounds (although I always check them to be sure).
Although I admit that this leaves the disadvantage of knowing Kanji from compounds but not what they mean by themselves when I'm not diligent enough to look up every part of a compound.
So that's it for my Kanji learning experience for now, maybe some others can shine some light on how they prefer to tackle Kanji and maybe some tricks to make the process more interesting. I've toyed with the idea of grouping Kanji within radical groups and learning them in batches that way but I've found to be more engaged in the process if I just add to my deck what I encounter although it might not be the most efficient way.
I've also used to use heisig but that one wasn't for me. I retain things better when I have a sound to associate with.
Would have quoted the whole thing but it's really long and this is the most relevant section. Anyway, learning kanji from compounds is not necessarily a bad thing. There are many, many cases where a certain character is virtually never used alone and there are 1 or 2 essential compounds (or names!) which are the only reason it's widely known. That's particularly why Heisig's approach is so valuable. If you focus on how to write them and a general meaning, you have the stroke count and general sense down, and then after that you start filling readings in mostly as you come across them in your actual Japanese reading and usage. And while learning the 音読み is good, it's a bad way to remember the character because so many are similar and sound the same, but you will find yourself remembering the character by the sound and forget what the words mean, or even worse you'll associate it with a single reading and forget that there are others which may be rather common as well. I much prefer seeing a compound word and knowing what it means but forgetting how to read it than being able to make a reasonable guess at the reading but having no clue what it means. You also have to be really wary of compounds that use 訓読み. It's really easy to look at a word like 船出 (to use a One Piece relevant vocab word :P ) and want to read it "senshutsu." But it's pronounced "funade."
Anyway how to best learn kanji has been a hot topic of debate for decades. In Japan it's basically rote memory and extreme repitition. To an extent for each person it will vary, but I think it's hard to deny the utility of building up characters from a slowly incrementing pool of radicals or smaller blocks which may not necessarily be radicals, hence Heisig's designation "primitives." If you can bear the procedure of learning the characters before knowing how to read them, it really is an astonishingly fast way to build your repertoire. And if you think about it, it's no good to memorize mass quantities of words if you're not going to be using and reading them, so in the reverse case it makes sense to fill them in as you go. Already having the foundation of "oh yeah I know that character" really helps put things in the bag. And for me personally, the 音読み come quite naturally without deliberate studying just by recognizing patterns.