As you may have seen from some other threads, I'm working with some people on a project where we're creating a Japanese keyword and example sentence for each of the kanji. Laxxy, Raulir and Darkshade are kindly getting involved so far.

I've had another idea in the back of my mind for a while though. A large chunk of joyo kanji simply aren't used much in daily life. What's more, there are a number of non-joyo kanji (RTK 3) which are used a lot. Would people like to help me create an RTK list that would exclude the kanji that are hardly ever used (吾、朋、旭 to name a few), and include some of the really common RTK 3 kanji?

Obviously it's not as simple as just removing the uncommon kanji and slotting in the RTK 3 kanji in the appropriate place. Simply by having lots of kanji that use the same primitive, you remember that primitive extremely well and if I take out too many kanji then I might be left with cases where a primitive only has one example kanji. This would reduce learning efficiency I think. So a balance would have to be made. Still, I think the massive 2000 number could be reduced to a figure that could be completed in a much shorter time period and have a real material impact on people's studies fast.

What do people think?

While I'm interested in joining in the first project you mention, the second one is less attractive to me personally. As you say, and I agree, Heisig takes you through the Jouyou kanji pretty well by primatives, and as you learn the primatives, some non-Jouyou kanji that are just a gimme if you know the primatives. Maybe some kanji ARE rare, but there's really not much of an effort wasted in picking them up and they "fit" there. But in terms of the first project, if they're that rare, then it may be difficult/impossible to find effective Japanese keywords and sentence examples for the rarely used kanji. The two projects are related, I can definitely see that.

I guess I still think it is "cool" to "know" a rare kanji or two, if it doesn't take me much skull sweat to rake it in.

My initial reaction was that this might be a good idea, but the more I thought about it the less attractive it seemed. I think for starters a lot of people (me included) will just want to know the [kana]jouyou[/kana] regardless of the usefulness or otherwise of some of the characters: not entirely rational, perhaps, but it would bug me.

I also think that a number of the rarely used ones aren't too tough anyway. Your examples 吾, 朋 & 旭 fall in to that category for me, and 朋 is a primitive that, for me, was worth the minutes it took to learn just to help remember 崩. And 旭 may not be much use in everyday reading, but it seems to occur in a lot of names (this is probably true of a number of the rarely-used [kana]jouyou[/kana]) and it would bug me not to have learned it, and who knows what might come up in the future that would cause me to regret leaving this and similar kanji out? I think there will be similar reasons for wanting to keep a lot of them, and then there's the time and effort you have to go to identify them and eliminate them from review materials. It just seems to me that it would be easier just to learn the buggers! I'm also kind of wary of modifying the Heisig system "to destruction".

I also thought 吾、朋、旭 was a strange choice of examples. 吾 serves as a primitive for other kanji (it's not part of joyo kanji by the way and was surely introduced because it's simple and serves as a part of other characters). 朋, too, serves as part of another important kanji, and I have seen it in actual names (朋香). This too is not a member of the joyo-kanji, but is very easy to learn. 旭 is also a character I see quite often (even though it's not part of the joyo kanji).

I can only fathom a few characters that aren't really worth learning early on (匁 comes to mind, and maybe 厘). I think going through the whole book is worth the effort.

JimmySeal Wrote:朋, too, serves as part of another important kanji, and I have seen it in actual names (朋香). This too is not a member of the joyo-kanji, but is very easy to learn. 旭 is also a character I see quite often (even though it's not part of the joyo kanji).

The examples wrightak gave are 人名用漢字 (じんめいようかんじ), kanji for names, of which Heisig says he included about 60. The 人名用漢字 at the time Heisig selected them were probably these (as at 1981):

丑 丞 乃 之 也 亘 亥 亦 亨 亮 伊 匡 卯 只 吾 呂 哉 嘉 圭 奈 宏 寅 巌 巳 庄 弘 弥 彦 悌 敦 昌 晃 晋 智 暢 朋桂 桐 楠 橘 欣 欽 毅 浩 淳 熊 爾 猪 玲 琢 瑞 睦 磯 祐 禄 禎 稔 穣 綾 惣 聡 肇 胤 艶 蔦 藤 蘭 虎 蝶 輔 辰 郁酉 錦 鎌 靖 須 馨 駒 鯉 鯛 鶴 鹿 麿 亀 佑 允 冴 喬 怜 旭 杏 梓 梢 梨 沙 渚 瑠 瞳 紗 紘 絢 翠 耶 芙 茜 藍 那 阿 隼 鮎 葵 伍 伶 侑 尭 孟 峻 嵩 嶺 巴 彬 惇 惟 慧 斐 旦 昂 李 栗 楓 槙 汐 洵 洸 渥 瑛 瑶 璃 甫 皓 眸 矩 碧 笹 緋 翔 脩 苑 茉 莉 萌 萩 蓉 蕗 虹 諒 赳 迪 遥 遼 霞 頌 駿 鳩 鷹

(The 人名用漢字 have been expanded enormously since then.)

I've definitely seen some of these in regular words (that is, not names), such as 虹 or 亀 or 呂. I think the RTK1 list is fine as it is.

I guess I should explain a bit further what I meant. Thanks for the input everyone.

I was never really suggesting that the list should be shortened and the less common kanji ignored. I just think they should be tackled at a later date.

As Heisig himself points out, the most important part of his method is the order in which he presents the kanji. The basic rule is to introduce primitive elements one at a time and make as many kanji as you can using the primitive elements you've introduced. He also takes care to leave it as long as possible before introducing extremely common primitives such as 人. However, before applying these principles, he had to choose which kanji to apply them to. He chose just over 2000 of them. The whole idea was that those would be all the kanji you would ever need. You could learn the writing and then the pronunciation afterwards with RTK 2.

Great idea... except along came RTK 3! So there are actually kanji that are really necessary that weren't in the RTK 1 list. If you follow Heisig's method then you'll have to study them after you've done all of RTK 2. So even now, the list is split in two with just over 2000 in the first bit and 1000 in the second.

I think that it would be better if you split things up earlier and did maybe 1000 kanji and then worked on their pronunciations, did another thousand and worked on them too. After that, you could do another 1000 if you were really keen and like Pangolin, it bugged you not knowing all of them.

I think the divide and conquer idea (of writing first then pronunciation) is in essence a good one but I think it's just been taken too far when the first step is wade through 2000 kanji. I remember thinking that it wouldn't end and it's a lot of work to get through before RTK starts having an impact on your Japanese studies.

I think it would be better for future learners if they could have a shorter list to deal with first and then another list once they've done that. I'm not suggesting that any kanji should be abandoned. I know that there are some people who burned through RTK 1 in 4 weeks but for most people it takes quite a while.

If I was starting again, I'd want to learn kanji using Heisig's logical ordering but I'd also want to learn the most common ones first. I don't think these two things are mutually exclusive and I think a balance can be reached between the two.

I know that the examples I gave are used in names quite a lot but I had forgotten that 朋 appears in 崩れる, that's a very good point.

You might ask why I'm bothered about all this if I've already done the RTK method anyway. I guess I'd just love to make it all as accessible as possible to future learners.

I have to agree that I think that RTK1 is a big chunk to swallow prior to RTK2 (and by the way, I'm not even really using my copy of RTK2). I've got right into RTK3, just continuing to chip away at learning how not to forget how to write the kanji. I guess I'm feeling that, what I'm doing now to learn readings could possibly have been done all along. Collecting sentences containing the words represented by the kanji that I've studied. But, I just can't say. I'm so atypical. I've been studying Japanese so long already, before RTK1, that probably what I'm finding as an easy recreational reading (albeit serious in the aspect of harvesting the readings) might be completely overwhelming to another paduwan. I know I completely tense up when confronted with a typical japanese webpage.

CharleyGarrett Wrote:I know I completely tense up when confronted with a typical japanese webpage.

Play it again, Sam! I think I would baulk at many Japanese websites if they were in English. I think their approach to "design" is partly at fault. There often seems to be a bewildering number of forms and junk like that. And zillions of links that seem to go around and around with virtually no content on each page. Maybe I exagerrate but I do find the layouts bewildering, on top of having to puzzle out the kanji.

wrightak Wrote:You might ask why I'm bothered about all this if I've already done the RTK method anyway. I guess I'd just love to make it all as accessible as possible to future learners.

I for one think this is a great idea. I would love being where I'm at in RTK1 (about 700 kanji), and feeling like I was almost done, instead of barely a third of the way. Heisig's critics often point out that his method doesn't help unless you go all the way (which I think isn't quite true, but a valid point) - this approach would basically eliminate that complaint.

So, great idea! If I can help with any of the logistics, let me know (I can't help with the language part yet, sadly!).

UPDATE:
I spent the afternoon working on trying to refine the idea. Using KanjiDic and the Heisig-Data file from Zig, I cross-referenced the 2500 most frequently used kanji against the Heisig word list.

It turns out that there's only one kanji in the top 1000 by frequency that isn't in RTK1 (2344:aim at). So making a High Speed RTK1 (HSRTK1) is really a matter of selectively taking kanji out of RTK 1 rather than inserting kanji from RTK3. [A HSRTK2 - the next 1500 - would be a more complicated mish mash of the two existing books].

Heisig RTK1 explicitly works from the premise that you learn a primitive, then learn all the kanji you can make in the top 2000 using all the primitives you know so far. As wrightak mentioned, the only change with High Speed RTK would be to change the number 2000 to the number 1000. The results? You'll have less practice with each primitive (which will make the primitives harder to remember), but you'll get through a useful set of kanji more quickly (and as Charley says, get on to the readings, if that's your bag).

The biggest drawback, though, is that this site would become less useful for review (and more awkward for study) - although there is always SuperMemo, etc., I suppose!

I put together an excel spreadsheet with what I came up with. Email if you are interested in seeing what I came up with. Next time I have a free day, I'll fool around with the kanji lists in the Easy RTK3 thread.

I am still not sure about the usefulness of this, but if someone does this, there are things other than frequency to consider, in particular,
-- all kanji that are themselves part of other kanji in the list have to be included
-- no primitive should be reduced to a single incidence (better take it out entirely if that is the case)

Also 'frequency' is not easy thing to define, I'd make it 'number of documents containing certain kanji in given representative selection'. To get texts representing everyday usage you can use selected forum or newspaper archives.

Assuming that the 1,000 highest frequency kanji that aboros checked against RTK I & III kanji is pretty much the same as the frequency data included in Jack Halpern's "Kanji Learner's Dicionary" then I'm not sure that 1,000 is the cut off I would make for a speed version of RTK. Sure these 1,000 kanji cover 95% of the kanji used in the Asahi newspaper for an entire year (the largest of any of the frequency studies to date), but just how good is even 95%.

95% means that when reading the Asahi newspaper, on average, you will not know one out of every twenty kanji you come across. However, kanji is a very information dense writing system, with on average 2 kanji per word. This means that out of roughly every 10 words you try to read you won't know at least one of the kanji used. So as good as 95% sounds you get bogged down quickly. (I know from experience... When I went to graduate school in translation years ago with only 1200 or so kanji firmly under my belt, I was always reaching for the electronic dictionary and kanji dictionary, etc. So much wasted time and frustration. I admit that the sheer volume of reading material and translation deadlines meant that info was coming in faster than my brain could remember and hold on to.)

I humbly submit that even a streamlined RTK shouldn't go as far as cutting the number to the most frequent 1,000. Better to go for the top 1,600 kanji which will give you 99% of all kanji used in the Asahi newspaper. (By the way, many Japanese language teachers consider the Asahi to be the most difficult newspaper for non-Japanese to read because it uses more kanji and compound words than the other newspapers do. So we were always encouraged to read it rather than the Yomiuri or the Mainichi.)

So, how far will 99% take you on average? It means that you won't know 1 kanji out of every 100 you come across, which boils down to not knowing one of the kanji in every 50 words or so. Much better than being stumped or having to guess at every 10 words.

If you do want to stick with the top 1,000 kanji, then I think it would also be wise to find kanji and reading textbooks that are graded, make sure that the kanji in your Speed RTK include the kanji presented in these texts. This way the student's initial reading material will stick pretty much to the learned kanji and they can build up kanji based vocabulary within this framework. Some good textbooks are "Kanji in Context" and "Basic Kanji Vol. I & II" along with "Intermediate Kanji Vol. I & II."

A number of good comments so far.

I'm using the Halpern frequency numbers from KANJIDIC because I couldn't find anything else in a format that was easy to use in Excel. I'm happy to use any other lists that seem reasonable and are available.

In order to try to capture all the most usable kanji, I took Halpern's top 1000, added the first 6 grades of [kana]jouyou[/kana] kanji and I'm in the process of adding the kanji identified in the RTK3: Easy Kanji thread and the kanji from
http://www.csse.monash.edu.au/~jwb/jouyoukanji.html
and
http://dict.risukun.com/HighFreqNonJouyou.htm

Of course, the problem with all this is that if I come up with a list that gets close to 2000, I'd probably want to pare down again.

Chadokoro_K Wrote:I humbly submit that even a streamlined RTK shouldn't go as far as cutting the number to the most frequent 1,000. Better to go for the top 1,600 kanji which will give you 99% of all kanji used in the Asahi newspaper.

Of course (as wrightak pointed out to me), the point of the HS RTK exercise isn't simply to put together a "top 1000" list. It's also to put together a HS RTK2 - covering maybe the top 2500 kanji. Basically, taking a new stab at the word list Heisig and Sienko settled on almost 15 years ago.

The debate over whether a shorter list is worth learning is, of course, well-explored. My argument is simple: 1) finishing a shorter list will give me a sense of accomplishment that will provide motivation for continuing my Japanese studies; 2) I'm not expecting to read the newspaper any time soon. Being able to read childrens books and some manga would be nice, though.

laxxy Wrote:-- all kanji that are themselves part of other kanji in the list have to be included
-- no primitive should be reduced to a single incidence (better take it out entirely if that is the case)

Well said, with the minor caveat that if an infrequent kanji is only used in other infrequent kanji, it can be postponed, and if it's used as a component only once or perhaps twice, it may be sufficient to use the underlying primitives in the more complicated kanji .

Unfortunately I don't know of any easy electronic lists of the high frequency data presented by Jack Halpern in Kanji Learner's Dictionary. White Rabbit Express (makers of some pretty awesome flashcards) provides cross-reference lists on their website. (http://www.whiterabbitpress.com/catalog/...-sp-3.html). But they only have cards through JLPT Level 2, so their cross-ref lists will only take you through the Basic Kanji textbook series. But I expect they will update this once their 3rd set of cards comes out (which will bring the number of kanji they present to around 2,000). So you may want to keep this link handy for future cross-ref possibilities.

You do make a very nice argument for the sense of accomplishment that can go with completing a shorter list. After my long absense from the world of kanji and Japanese I am now at #925, and so yes, especially if it were my first time through, it would feel a lot better to be finishing at 1000 rather than be only at the half-way point.

I may reverse my earlier position and argue that such a list could create a solid base for beginners to start reading and begin to learn how kanji function in the real world. There's nothing that says that they can't begin reading and using this base set while following it up with round two of the Heisig approach to another 800 or so kanji. (Making sure to incorporate high frequency words. Again, my ideal choice would be to use the kanji that would take you to 99% coverage.) Then the learner could begin to explore those kanji in the real world as they begin a third RTK-approach list, etc. This would add the element of built-in review of the primitives and kanji that function to buildup other kanji.

I think you're right on target to include grade 1-6 kanji if you want to be able to read kids books and manga. Kids books will use kanji for animals, for example -- kanji that aren't commonly seen in newspaper columns or in magazines but of course everyone knows them because they are easy words.

I still think high quality textbooks together with manga (which provide easily understood context as well as natural language) are the way to go for learning reading and how kanji operate in the real world. Especially if anything short of the full Joyo-list is taught as the base from which to begin reading.

Of the texts I mentioned earlier, "Basic Kanji" and "Intermediate Kanji" offer lots of writing practice for words (not simply the copy this kanji 100 times approach) and they lead the student through grouping readings like Heisig does in RTK2. At this stage, graduated reading material provide a further boost to self-confidence. (It just plain feels good to be familiar with the majority of whatever is written on a page.)

I think I misundersood your original goal. I applaud your efforts to reinvent the kanji list that Heisig and Sienko came up with all those years ago. It will be very interesting to see the list you come up with. Or perhaps lists, if you like the idea of taking a learner through two or three graduated lists using the RTK method to get them to a total of say 2,500 kanji after which they can contnue a lifetime of acquiring new kanji as they meet them.

Personally, I want to go through RTK III before switching to the "learn new kanji as you encounter them" approach. But I'm a bit put off by the fact that I can probably do without learning a good chunk of it right away. So I would welcome a list of 2,500 really useful kanji over the present selection in RTK I & III.

To be honest, I don't think the frequency lists cut it that well although they are a source of input. I believe that they just contain frequency data from newspapers and newspaper language is very formal and wouldn't be useful to a Japanese learner. After completing the first 1000, I would hope to put a learner in a position where they could WRITE lots of sentences in Japanese and be productive. I wouldn't want to try and put them in a position where they could understand everything (it won't happen). Although this seems to be what everyone who gets involved with RTK wants. Focus always seems to be overly placed on reception skills rather than productive ones.

I was planning on using judgement and getting feedback from Japanese people. The idea is just to get something more useful, it doesn't have to perfect.

I'm also afraid that I don't understand the above argument of 1000 kanji only covering 95% of kanji being a bad thing. Surely if you wanted to cover 99% or above, you'd just learn the whole of RTK1&3 and this whole high speed idea would be pointless!

I was thinking of sorting the RTK1 kanji by frequency, and learning them in that order. Of course, you still need to finish them all to read a paper etc, but knowing the more common ones earlier would help with more casual reading and reinforcement from exposure to kanji in the wild. The list needs to include all primitives before any kanji that contain them, so it would be heavily weighted with primitives at the start. It also wouldn't follow Heisig's method of just adding a single primitive and learning all the kanji that can be produced at that point, and you wouldn't be able to use this site until the add-new-kanji function is extended. To perform this sort I'd first need to know the primitives that make up each kanji, see http://forum.koohii.com/showthread.php?pid=2941#pid2941 for more.

That "in order of frequency" is a wheel that has already been invented, and (IMO) proven to be flat...well, at least not as round as Heisig's method. Of course, if you want to do that, then go ahead. I've been studying kanji a long time before Heisig, and it didn't work for me. Part of the wonderfulness of Heisig's method is the pure common sense of the order that the kanji are studied. (ie, learn the primatives first, learn the kanji that are built from that primative. Not the way they would come out with kanjis prior to the primatives that build them and then the primatives all gocha gocha). Well, I admit that "in order of frequency" is common sense too. It's just that 5 kanji related to the same primative have a more cohesive set of continuity than frequencies 5th, 6th, 7th, 8th and 9th. Naturally, once they're learned, you don't need to recall them in Heisig frame sequence, or in frequency of appearance in newspaper compounds either. After learning them, you'd have more chances to review the first ones that you study. That's about the only advantage that I can see.

I'm a convert to this method, and I'm not likely to be easily persuaded that it's bad or that these "ituitive" alternatives are going to be improvements. I realize that people are different, so YMMV. That's cool.

CharleyGarrett Wrote:Not the way they would come out with kanjis prior to the primatives that build them and then the primatives all gocha gocha

That's the reason I need the info on what primitives make up each kanji, so they can be included in the list before any kanji that uses them.

CharleyGarrett Wrote:After learning them, you'd have more chances to review the first ones that you study. That's about the only advantage that I can see.

Yep, that's the advantage, to review them and to make use of them. The disadvantages I mentioned above. I don't know if the advantages outweigh the disadvantages, but there's a chance they might, and I'm interested in building the list regardless just to see how it turns out, if not actualy to use it.

Chadokoro_K Wrote:White Rabbit Press (makers of some pretty awesome flashcards) provides cross-reference lists on their website.

Thanks for the good word!

We've moved the cross-reference files to the downloads section of our new support site:
http://www.whiterabbitpress.com/support/

Also we have an online tool "Kanji Finder" which you can use to generate your own cross-reference files. Details here:

http://www.whiterabbitpress.com/support/...ticleid=26

Max Hodges
Publisher
White Rabbit Press
http://www.whiterabbitpress.com