As the title says. What I generally do, using mnemosyne, is look at the keyword. If I can remember the kanji with only the keyword, I rate 4 or 5, depending on the time. If i need the story, I rate 2 or 3, again, depending on the time. If I don't remember it, or I get a part wrong, I rate 0-1, aka total fail. Problem is, I never write them down when reviewing. Is this bad? Is my method ok? What method do you guys follow?

Thanks : )

[ファブリス]

I think you should always write down the character, at the very least "draw" with a finger on the palm of the hand. Or see the writing in your mind's eye (ie. visualize it) if you're reviewing on your mobile phone, cramped up in a Tokyo train.

I just about to make this same post today. Coincidental.

I was gonna ask what people do when they review with something like Anki. Example, do they just remember the answer, and press a button they deem appropriate, do they write down the answer even if they got it right, or do they say the answer out loud to themselves.

And what do they do when they fail a card? Do they just keep going until all fails are correct, do they redo what they did to learn it in the first place?

And I was gonna ask them what their general card fail rate was per review.

I use anki.

I write down the Kanji regardless if i know it or not.
I press good if i got it correct 'Again' if it's even slightly wrong.(didn't connect stroke, etc.)

I always complete failed cards and I'll reread the story when i fail it.

I pass 85% of cards on average.

KMDES Wrote:I just about to make this same post today. Coincidental.

I was gonna ask what people do when they review with something like Anki. Example, do they just remember the answer, and press a button they deem appropriate, do they write down the answer even if they got it right, or do they say the answer out loud to themselves.

And what do they do when they fail a card? Do they just keep going until all fails are correct, do they redo what they did to learn it in the first place?

And I was gonna ask them what their general card fail rate was per review.

Usual one rates each card appropriately even if it's a mature card. You fail it if you can't read any of it write/have no clue of what it means. You usual rate it hard if you can only remember 2/3 of the card. You rate it good when you can read/get the whole card and rate it easy when you can read it fast/have no problems at all.
This is for recongition type cards though.No writing.
I usual have a separate deck for writing

ta12121 Wrote:Usual one rates each card appropriately even if it's a mature card. You fail it if you can't read any of it write/have no clue of what it means. You usual rate it hard if you can only remember 2/3 of the card. You rate it good when you can read/get the whole card and rate it easy when you can read it fast/have no problems at all.
This is for recongition type cards though.No writing.
I usual have a separate deck for writing

I'm not asking how people generaly do it as that's pretty obvious, I'm asking how each individual actually goes about it as most people don't do it the same way. Basically, I'm not looking for insructions, I'm looking for method versus performance.

KMDES Wrote:

ta12121 Wrote:Usual one rates each card appropriately even if it's a mature card. You fail it if you can't read any of it write/have no clue of what it means. You usual rate it hard if you can only remember 2/3 of the card. You rate it good when you can read/get the whole card and rate it easy when you can read it fast/have no problems at all.
This is for recongition type cards though.No writing.
I usual have a separate deck for writing

I'm not asking how people generaly do it as that's pretty obvious, I'm asking how each individual actually goes about it as most people don't do it the same way. Basically, I'm not looking for insructions, I'm looking for method versus performance.

In terms of srsing split it among my decks for what I'm really testing about. So my sentence deck=context only while my vocab deck is for vocab only. Both purely recognition of course, otherwise it take me almost the whole day to do.
I write down my reviews for my production deck. it contains kana to kanji vocab cards and RTK keywords as well as some small sentences in there.

The way I usual do it is split it up and work on it that way. I only have 3 big decks as oppose to small ones spread out. It's just more time-efficient that way/easier to organize.

Hope this helps in some way.

KMDES Wrote:I just about to make this same post today. Coincidental.

I was gonna ask what people do when they review with something like Anki. Example, do they just remember the answer, and press a button they deem appropriate, do they write down the answer even if they got it right, or do they say the answer out loud to themselves.

And what do they do when they fail a card? Do they just keep going until all fails are correct, do they redo what they did to learn it in the first place?

And I was gonna ask them what their general card fail rate was per review.

I have two decks - vocab and grammar deck. For vocab, if I can't read the kanji word (or kana word) and remember the definition, I fail it. I have a max of 10 failed cards at a time, and it cycles through until I remember it. If I'm really having trouble getting something to stick, I read it out loud a couple times and then move on. I don't write anything down since I tend to do my reviews at night in bed when my dog doesn't care what language I'm rambling in.

I learn directly through SRS - I pull words through Yomichan and learn them as they come. Grammar is different - I read the DOBJG section, then unsuspend the cards.

Erm...right now, my general fail rate for vocab is pretty high since every vocab card (99%) is new vocab for me, and since I do new cards every day I have a high fail rate for some because it takes them a while to stick (5 days or so?). My grammar is pretty high (90ish %?), since I lovelove grammar and I tend to remember it easier, though I do get tripped up.

I tend to hit "hard" a lot, so it leaves with me a lot of reviews (only 450 cards in the deck [not all used yet], and right now I'm averaging about 100 reviews a day or so, and that's just the expired cards) because I tend to hit "hard" even if I got the card right if I feel a little uneasy about it or think I probably won't remember it in a few days. I rarely (if ever) hit 4, or "superdooper easy". Maybe if it was a number. Either that or like "hi".

I think that's what you were looking for? If not, feel happy to point me back to the right track.

I review RtK using the shared deck with anki. I look at the keyword and write the kanji down, if I got it correct I rate good if I was able to draw it without thinking I rate it easy or very easy. If I get the strokes even slightly wrong I rate it again, but I'm masochistic like that.

Anyone have the kanji as the question, and the meaning as the answer? That's how I've been studying RtK1 this whole time (currently at 700) and now I'm afraid if this is a bad way! -_-'

joeschmo88 Wrote:Anyone have the kanji as the question, and the meaning as the answer? That's how I've been studying RtK1 this whole time (currently at 700) and now I'm afraid if this is a bad way! -_-'

Heisig himself and most people on here suggests that you review keyword -> kanji, so that you focus on trying to remember the actual kanji and not just see the kanji and memorize its keyword. If you want to see discussions of this you can just search the forums.

Are you using Anki? If so, it's easy to flip all your cards if you want to. Go to 'browse items' and then the 'card layout' option and just click 'flip'.

TwoMoreCharacters Wrote:Heisig himself and most people on here suggests that you review keyword -> kanji, so that you focus on trying to remember the actual kanji and not just see the kanji and memorize its keyword. If you want to see discussions of this you can just search the forums.

Are you using Anki? If so, it's easy to flip all your cards if you want to. Go to 'browse items' and then the 'card layout' option and just click 'flip'.

Oh man, did I just screw myself up in actually learning to write kanji? After searching/reading about the whole debate between Kanji --> Meaning vs Meaning --> Kanji, I feel totally down now. It really is a split between whether or not you want to recognize kanji easily vs writing kanji easily. I don't want to start a discussion on this issue since its already been done, but now that I'm already 700 kanji in, do you think I should just continue or switch? I tested out the switch yesterday, and yes, I admit it is quite difficult to write it out. I feel like I should just continue, as writing will come along naturally later on. Same thing goes for the majority that go Meaning --> Kanji, recognition will come along naturally. Ahhhhgghhh I'm only mad at myself! Sorry for the vent!

Writing won't come along naturally. My Japanese teacher last semester couldn't write Kanji for some things like 髪　but I can just fine. If you can write it you can most likely read/recognize it. But not necessarily the other way around. Thats why computers are so good you don't need to write anymore but yeah I'd switch if i were you. You're less than half way.

my reviews seem to be taking longer lately, and I haven't been able to add new cards like I'd like to... feeling pretty down about it tbh, like I'm losing my pace. I guess it's because uni started again, but seriously doing 100 cards feels like it takes LONGER now than it did a month ago. No idea why hehe.

joeschmo88 Wrote:Oh man, did I just screw myself up in actually learning to write kanji? After searching/reading about the whole debate between Kanji --> Meaning vs Meaning --> Kanji, I feel totally down now. It really is a split between whether or not you want to recognize kanji easily vs writing kanji easily. I don't want to start a discussion on this issue since its already been done, but now that I'm already 700 kanji in, do you think I should just continue or switch? I tested out the switch yesterday, and yes, I admit it is quite difficult to write it out. I feel like I should just continue, as writing will come along naturally later on. Same thing goes for the majority that go Meaning --> Kanji, recognition will come along naturally. Ahhhhgghhh I'm only mad at myself! Sorry for the vent!

Well, it could be worse, you've only done like one third of the book.
I think you should switch. No need to feel down, think it as reversing cards: you've trained for one exercice, and now you're training for another. And despite the fact it's a bit difficult to produce kanji at first, you're already acquainted with 700 kanji, so you should re-learn them quite quickly.
Plus I find the first 500 kanji to be easy compared to the rest, so you really should switch now while there is still time.
Good luck!

joeschmo88 Wrote:Oh man, did I just screw myself up in actually learning to write kanji?

I just wonder where you got this idea from. AJATT, maybe? *whistle* The owner is suggesting a method to make reviews easier, by putting the kanji and story on the front. Story on the front would be fine, but hidden, Kanji however is a no go. So it's better to start over, or continue but this time with keyword -> kanji.

What you will want to do is to write down kanji at least once, when you are first learning them, and when you are reviewing them. When you write them down while you are reviewing is, that you can actually see if you made a mistake, and rate your performance accordingly. You mustn't go hardcore and rate a card wrong, if you make slightly longer strokes, or too short ones, as long as the kanji in question resembles the original as closely as possible. You can set the rules how to grade yourself, by yourself, depending on what you are counting as error, hard, good or easy.

And as EratiK said, what you have done could have been worse, look at his example and don't ever repeat it. *click*

I just finished doing RTK1 deck on Anki, and this is pretty much how I did it.

If I can remember and write down the kanji (from memory. don't look up to the story you wrote down) in few (about 5) seconds = EASY.
If less than 15 seconds = GOOD.
If more than 15 seconds = HARD.
If it takes more than 30 seconds to remember and write it down, even if I got it correct = FAIL
And of course if it's wrong (even a little dot) = FAIL

I almost never had to look up to the stories, which I had written down, kinda just remembered them when I failed and looked at the kanji afterwards. Oh and at the end my total correct rate was being 93% or so and mature rating 89% something.

darkmind35 Wrote:I just finished doing RTK1 deck on Anki, and this is pretty much how I did it.

If I can remember and write down the kanji (from memory. don't look up to the story you wrote down) in few (about 5) seconds = EASY.
If less than 15 seconds = GOOD.
If more than 15 seconds = HARD.
If it takes more than 30 seconds to remember and write it down, even if I got it correct = FAIL
And of course if it's wrong (even a little dot) = FAIL

I almost never had to look up to the stories, which I had written down, kinda just remembered them when I failed and looked at the kanji afterwards. Oh and at the end my total correct rate was being 93% or so and mature rating 89% something.

Good advice that might work for some, or many, people.
Personally, I stopped rating my reviews (easy,hard) based on the time it takes me to recall the kanji, and now just hit spacebar if I can remember it within a reasonable amount of time. The extra step of rating my recall time just made the reviews too tedious and slow for me and in the end it'll all settle thanks to the SRS system, I think. I felt the exact intervals don't matter anyway (SRS is just statistical, the times aren't exact) so I might as well give a positive boost to my review speed this way.

Nagareboshi Wrote:I just wonder where you got this idea from. AJATT, maybe? *whistle* The owner is suggesting a method to make reviews easier, by putting the kanji and story on the front. Story on the front would be fine, but hidden, Kanji however is a no go. So it's better to start over, or continue but this time with keyword -> kanji.

Does anyone have any scientific proof besides 'Heisig said so' that it's better to do keyword->kanji over kanji->keyword? Or anyother proof besides that's what everyone else is doing?

The proof comes from experience. If you have taken classes you know this too well. There are kanji you can read but when it comes time to write it your completely stumped. So I use to be able to look at 推薦状 and be fine to read but try to get any of my classmates to write and they would just look at you with blank faces. I don't need a grad school kid in a lab to tell me this works better in my experience. You may be different and that's great but find out for yourself and that's worth more than what anyone else can tell you.

NoSleepTilFluent Wrote:The proof comes from experience. If you have taken classes you know this too well. There are kanji you can read but when it comes time to write it your completely stumped. So I use to be able to look at 推薦状 and be fine to read but try to get any of my classmates to write and they would just look at you with blank faces. I don't need a grad school kid in a lab to tell me this works better in my experience. You may be different and that's great but find out for yourself and that's worth more than what anyone else can tell you.

That's a completely different problem entirely. The reason you can't write is the same reason we're having the conversation in here. It's lack of muscle memory and other encodings in the brain as visual encoding doesn't equal tactile encodings. These encoding can be implimened via kanji->keyword studying easily.

There's reasons behind how things work and don't work. 'Just because' doesn't cut it anymore.

I said above that when you go keyword > kanji, you focus less on the keyword - which you'll eventually want to be phasing out/replacing with the more accurate meanings (if there are any) anyway - and you're trying to produce the kanji from your mind. If you go kanji > keyword, you just see and recognize the character while you work your brain into remembering the keyword - something that's less important. You're putting the focus on the wrong aspect.

These "encodings" you're talking about will just as well be implemented by practicing and becoming better at remembering the components the character is made of, and if you write the kanji down as you review them I don't see how it would be any inferior to going kanji > keyword.

And again, there has been tons of threads on this.

TwoMoreCharacters Wrote:I said above that when you go keyword > kanji, you focus less on the keyword - which you'll eventually want to be phasing out/replacing with the more accurate meanings (if there are any) anyway - and you're trying to produce the kanji from your mind. If you go kanji > keyword, you just see and recognize the character while you work your brain into remembering the keyword - something that's less important. You're putting the focus on the wrong aspect.

These "encodings" you're talking about will just as well be implemented by practicing and becoming better at remembering the components the character is made of, and if you write the kanji down as you review them I don't see how it would be any inferior to going kanji > keyword.

And again, there has been tons of threads on this.

I can see how focusing on reconstructing the kanji in your head can be helpful. It also could be done kanji->keyword by reconstructing it after the answer by averting your attention from seeing the kanji and maybe writing it on paper or in your head. Of course you are training your brain to remember the kanji because of the keyword, which you never see in the wild in Japanese as opposed to kanji->keyword which you could use in the wild to stumble your way through kanji (some of those compunds aren't gonna make any sense what so ever though). This is why I'm looking for data that supports either one (I personally did RTK twice keyword->kanji).

Encodings are what we use to remember basically. Here's the wiki entry. Encodings aren't always transferable either, else someone would be able to do something just from watching someone do it.

KMDES Wrote:

Nagareboshi Wrote:I just wonder where you got this idea from. AJATT, maybe? *whistle* The owner is suggesting a method to make reviews easier, by putting the kanji and story on the front. Story on the front would be fine, but hidden, Kanji however is a no go. So it's better to start over, or continue but this time with keyword -> kanji.

Does anyone have any scientific proof besides 'Heisig said so' that it's better to do keyword->kanji over kanji->keyword? Or anyother proof besides that's what everyone else is doing?

Going keyword -> kanji is just one step in the overall process. And it does not need scientific proof to tell why this is the better (not the only way) to do it. As you should know yourself, writing is not hard at all, the way the method works. But what's the use to constantly see the kanji, copying it off, seeing the keyword after pressing a button, and move on? It is incredibly easy, anyone can do it, no problem.

But what about the later stages of learning? The keyword is supposed to fall away, as well as the story, being replaced by ひらがな and かたかな, to produce words and whole sentences, containing kanji. How would this work, then? Does he go sentence on the front, just copying it off by hand, without any effort involved to remember the reading of the sentence, written on the back of the card? And what happens when time comes to move away from SRS altogether? There is only memory -> reading -> kanji or memory -> kanji -> reading. Only going one way isn't the way to go then. But not doing it the other way around, involving the need to produce the kanji from memory, which in this case cleary does not happen, is a waste of time.

If he has done it like this up to this point, it mustn't be wrong per say, because he should know the keywords pretty well, and mayhaps he is able to do it the other way around. Keyword -> Kanji. But with RTK having random keywords, not matching the real meaning at all in some cases, there is no need to go kanji -> keyword.

IF being able to write is his goal at all, he should at least try doing it the other way around, keyword, story if needed, kanji. If he only want's to familiarize himself with them, however, to be able to read later on, then this might be a way to go about it.

@KMDES

I was under the impression that active recall is well-established as superior to passive recognition in the literature since the middle of the 20th century or so, hence in the case of Heisig you use the keyword to cue the story, neither of which (keyword and story) are the targets for the SRS reviews, the target is to produce the kanji, to internalize the kanji as instantly recognizable icons which emerges from a bottom-up process. You want to reinforce those encodings in a full and rigorous way, including initiating those internal motor sequences, whose encoded neural paths are triggered upon simply viewing a kanji and which thus would also reduce the utility of motor memory for recall as it's also being given a handicap in kanji→keyword. Keeping in mind that active recall/writing → most effective for recognition/reading.

I think we've made well-reasoned arguments for years about why to do keyword→kanji that has rarely relied upon saying ‘Heisig said so’; even without scientific research into the Heisig method I think it's clear that kanji→keyword is absolutely incorrect at worst, and wasteful at best.

Edit: On a related note, we've discussed this research before, but here's a recent paper on active recall (active learning/retrieval, really) that accepts the importance of elaborative study/encoding but also stresses retrieval as an equally or even more important active recall process that is part of a dynamic learning paradigm where it's essential to continue to be active upon each subsequent review/cued reconstruction (which fits well with my feeling that the SRS should be considered a flexible HUD rather than a static vessel): http://www.sciencemag.org/content/331/6018/772.full - Here's another paper that looks at how elaborative encoding and retrieval interact as well as the superiority of ‘weak’ cues for promoting retention.

Related: http://www.sciencedaily.com/releases/201...153301.htm

Also, another paper by Karpicke, et al. - Spaced Retrieval: Absolute Spacing Enhances Learning Regardless of Relative Spacing - This and other papers on expanding spaced retrieval by Karpicke, et al. fit well with my comments that the algorithm isn't as important as spacing itself in conjunction with user feedback for scheduling (and also there's finding that ideal delay for initial reps), where increasing space is almost incidental.