Hi everyone,

I'm a little bit confused about the pure groups introduced in Chapter 2 of RTK2. It is my understanding that for these subset of kanji, each time we see a particular signal primitive, we will automatically know what the Chinese reading will be. For example, if we recognize the Chinese reading of 中 to be ちゅう, then we automatically know that 仲, 沖, and 忠 also share the same Chinese reading ちゅう.


Now, why is it the case that for the primitive 白, which in frame R-89 is given to have the Chinese reading はく, we have that 習 is read しゅう (see frame R-420)? I thought that since this character contains the primitive element 白, the Chinese reading must be the same, e.g. 習慣 must be pronounced "はくかん", when it is really pronounced "しゅうかん".



Any help is much appreciated!

Hm, if you reread the beginning of the pure group chapter carefully, he doesn't say all other kanji exhaustively will be read ちゅう if they have 中 (even if practically it works, but 中 is probably an exception in that respect), he says "in volume 1, only 3 kanji will match the definition (of a pure group) for 中/ちゅう ". And that's what a grouping is, an arbitrary gathering that obey certain laws. So even if 習　has 　白, it's not part of the 白/はく pure group,

Lightninjaj Wrote:... must be pronounced "はくかん" ...

Except it would be pronounced はっかん.

(2016-02-15, 8:03 pm)Lightninjaj Wrote: Hi everyone,

I'm a little bit confused about the pure groups introduced in Chapter 2 of RTK2. It is my understanding that for these subset of kanji, each time we see a particular signal primitive, we will automatically know what the Chinese reading will be.

That's a misunderstanding of what he's saying. The pure groups are a method of grouping kanji to learn their readings all at once, *not* a method for predicting readings of kanji you haven't seen. Using the method to predict readings requires a deeper understanding of how kanji were formed and etymological information about the individual characters.

No, it really is supposed to be just* that (for the kanji in RTK1, that is). Heisig even says to be "very careful to note the special conditions that occasionally accompany the signal primitives" but omits the most basic one: many components, like 白, are only phonetic when they're on the right.

*With the caveat that "pure" doesn't mean the kanji can't have multiple on'yomi, or that the pure reading is the most common one for all of them. For perfect groups (pure + unique reading) and a computer-generated list of phonetic components, see this thread.

(2016-02-16, 10:56 am)yudantaiteki Wrote: Using the method to predict readings requires a deeper understanding of how kanji were formed and etymological information about the individual characters.

My script didn't understand anything, and it still managed to find a lot of pure and perfect groups. More than Heisig did, in fact. (白 is not one of those groups - I didn't have component positioning information).

Thanks alot for the help guys!

I'm leaning towards Vempele's explanation. In the beginning of Chapter 2, Heisig states "... each time you see this primitive element [referring to 中], you can be sure what the Chinese reading will be." He then warns the reader to "be careful to note the special conditions that occasionally accompany the signal primitives". Warning us about exceptions to these conditions seems to imply that we're supposed to extrapolate the rules to new, unseen kanji when applicable.

...Come to think of it, the pure groups were also supposed to be *complete* lists of kanji containing the component, so any unseen kanji should have been non-RTK in the first place. Many but not all pure groups remain pure for the bigger kanji list that I used (all 4000+ kanji that appear in EDICT).

(2016-02-15, 8:03 pm)Lightninjaj Wrote: I'm a little bit confused about the pure groups introduced in Chapter 2 of RTK2. It is my understanding that for these subset of kanji, each time we see a particular signal primitive, we will automatically know what the Chinese reading will be. For example, if we recognize the Chinese reading of 中 to be ちゅう, then we automatically know that 仲, 沖, and 忠 also share the same Chinese reading ちゅう.


Now, why is it the case that for the primitive 白, which in frame R-89 is given to have the Chinese reading はく, we have that 習 is read しゅう (see frame R-420)? I thought that since this character contains the primitive element 白, the Chinese reading must be the same, e.g. 習慣 must be pronounced "はくかん", when it is really pronounced "しゅうかん".

There are more cases like this. First of all many of the "pure" groups are not pure anymore if you leave the area of the (old) RTK1. Second many of the RTK2 groups have ommitted members which are introduced later in the book.

In the case of 白 it is not only 習. Vempele wrote, it is not はく because 白 is not on the right, but there are also these two which are stand alone and on the right:
白 => びゃく 黒白 (こくびゃく) → black and white; good and bad
拍 => ひょう 拍子 (ひょうし) → beat; rhythm

Independent of the wording "pure", it is still helpful to learn 白 / 伯 / 拍 / 舶 / 泊 / 迫 first together as はく/ぱく (and exceptions later on).

Here's a list of truly "pure" phonetics and kanji (limited to phonetics where there are at least 3 kanji with it): http://pastebin.com/j7yFasK0

Might not be truly pure once you take non-joyo kanji into account.

I believe "pure" groups have some other definition other than you can predict with 100% certainty the reading of a kanji which contains the signal primitive.  Something along the lines of ..every kanji that contains this component sometimes has the same reading.

There was some discussion about "perfect" groups which seem to be what you are looking for.  And you can find them in the very useful list that vempele created.  I believe there are 325 perfect jouyou kanji - not very many.  The perfect kanji would be the ones that have 100% in both the "reading coverage" and "kanji/primitive coverage" columns.

There is a larger group (769 perfect and non-perfect jouyou kanji) which probably doesn't have a name, but have the same reading 100% of the time.  These would have 100% "reading coverage" but not necessarily 100% "kanji/primitive coverage".  So if you were to memorize the readings of these 769 kanji, you would be certain of the readings of over 1/3 of the jouyou kanji.  This is the strategy I will be taking in about a month after I finally finish rtk.