I'm curious what recall rate others find acceptable when doing reviews. Do you remember 80 per cent, 90 per cent, 95 per cent? Mine varies, of course, depending on whether I'm having a good day or an off day, but overall, with this method of learning the kanji, I'm thinking that anything under 90 per cent might mean you're not giving enough time for the image to solidify in your mind before moving on to the next kanji. Thoughts?

I hovered around 80% throughout, some days better, some worse. Since I finished, my rates improved, except on the occasional day I can't be asked and just failed a ton. These cases are when the story's on the tip of my memroy, but...

I would be reluctant to say recall<90% means I and others like me suck, because part of the point of self-directed learning is to figure out what your individual parameters are and adjust accordingly.

I think the daily recall rate is not so important. It'll reflect a lot of newly learned kanji for which its normal to have a lower recall rate, but should increase over 1 to 2 weeks of reviewing.

For myself, I look at the recall rate of long term kanji rather than the short.

My Anki stats:

Correct over a week: 95.2% (492 of 517)
Correct under a week: 88.3% (3550 of 4020)
Correct unseen: 86.6% (688 of 794)

The longer I've been studying the more the recall rate improves. Note that the way Anki categorises "over a week" doesn't mean exactly over one week, its more like three or four in practice.

I have been adding info on my daily previews and added cards to a spread sheet for about one and half months now and my average recall rate over that time is about 95%. The daily recall rate is usually something between 90% and 100% but I have had bad days too.

I think that, as long as the average stays at about 95%, I'm good but if it goes below 90%, I'll slow down and start looking for what I'm doing wrong.

The first half of the book I felt if I wasn't getting 90%, I wasn't spending enough time. But then I figured if I didn't move faster I wouldn't finish the book. So the second half of the book my recall rate for blue stack was about 50% and for orange stack probably about 70% (depending on the day).
If I would have tried to take the time to learn really well the kanji (and thus have better recall rates) I don't think I would have finished. So to me, what was far more important than having an acceptable recall rate, was have an acceptable number of Kanji learned in a day.

Generally 80-95% is best. The effort vs reward curve is not linear, and the jump from 80->90% may take half the effort it takes to jump from 90-95%. The goal is not perfect recall - if you fail something you can study it again. The goal is the most efficient use of your time. 90% is nice if you have it, but I wouldn't worry if your recall is less. Less than 80-85% and you may want to start doing something differently

vosmiura Wrote:Note that the way Anki categorises "over a week" doesn't mean exactly over one week, its more like three or four in practice.

That's fixed in the development tree :-) The explanation is better now

I've been having problems with this as I've been trying to go at a pace of 50/day now. The old cards are fine, it's just when it comes for that day 3 review it is hell.

So now I'm doing a review system before I add a set of cards and test them. Whenever I finish my story with a kanji, I add the keyword on one side of a flashcard (hand-made and small, I cut 5x3s into fourths) and the kanji in small writing on the other side. I don't write my story on the card (though I write all my stories down somewhere). That's too time consuming.

Then my schedule goes something like this (keep in mind I'm still learning 50 new ones while doing this)
Day 1: Review the set I made yesterday at least 2-3 times depending on how well I remember it. Basically, review it as much as I can until its grained into me.
Day 2: Review in the morning when I wake up. Correct any problems I have at the time. Review once more at night.
Day 3: Do not review the stack of cards in day 3
Day 4: Review at night once, and if I have any problems, correct them. Add the cards to RevTK and test them the next day.

I just started this a few days ago and have had a 2 day haitus so haven't be able to fully follow through with it. But my cards that I've been funneling through the system have been very memorable now. How this will play when I go a week without review, I'm not sure. But it can't be any worse than not reviewing them more before you add I suppose.

It also keeps what I've been doing a lot more organized and will force me to add cards to the site consistently. Before I would go at 50 a day and sometimes just not add them on the site until after 2 days later (without having reviewed them either) and then I have 150 kanji I made stories for but haven't added to the site...

My personal pass mark was 90%, and I wasn't really happy unless I got 95%. By that I mean getting every stroke correct. I took seriously Heisig's advice about not rushing, believing that this would be more efficient in the long run. Studying on average 1-2 hours a day, I got through RTK in 20 weeks.
While I'm certain that it is possible to go a lot faster if you have time to do 5 plus hours a day, I'm quite sceptical about the benefits of rushing through RTK while only scoring 85% or less on your reviews. I certainly don't think you can say you have completed RTK if you are getting anything less than this in your reviews.
I believe Heisig wrote somewhere that you should spend a few minutes letting the image for each kanji really settle into your mind. This is perhaps the best advice he gives.

@naniwa

I agree. I tried forcing my way through the book fast, but it wasn't for me. In the long run, it only took all the fun from it when I had lots of other stuff to do.

Now I'm taking it slowly and the only absolute must is to review every day and keep the red stack to a minimum. Sometimes it feels that I'm not going anywhere, but all the cards I add, I'll keep.
I think, I'll pick up some speed during the Christmas holidays when I don't have to worry about school..

Anyway, I think it's better to get the kanji solid in your head, instead of having them in your red stack..

I was always happy if my rate was above 80%. I guess my average would be in the high 80s though.

I find that I recognise a lot of kanji when I see them and know what they mean, but can't think of the specific key word. I can remember my story, how it relates to the kanji parts and the meaning of the kanji, but not the word itself. I don't let this worry me as ultimately the aim is to drop the English and learn Japanese readings for them. I guess my brain is already starting that process by just keeping the meaning of the kanji...

This also means that when I'm reviewing on this site, I often write down two or three kanji when I see a keyword. Each one has similar meanings and if one of them is correct then I'm happy.

Naniwa, may I take exception to your stated view, or at least how you said it?

naniwa Wrote:I'm quite sceptical about the benefits of rushing through RTK while only scoring 85% or less on your reviews. I certainly don't think you can say you have completed RTK if you are getting anything less than this in your reviews.

If that is your standard for yourself, fine, but by your high standard's I'm just fooling myself, and the 1500 cards in my fourth stack are worthless. But I guess that's why I'm doing self-directed learning, where I am best judge of my own progress, right?

Not worthless, by any means, Billyclyde, but I think in the long run if your goal is to learn all the kanji, then it might be better to slow down a little and get your percentage up a bit higher.
The process of learning kanji is virtually never-ending, my own recall rate is certainly not 100%. But I see many times on this forum, people writing about trying to rush through the book in record pace, and I must say I think that can often be counterproductive if the recall rate isn't high enough.
People can make their own judgements about what is acceptable, but for me at least, a goal of 90% seemed like a good target.

Naniwa, my main point is that I bristled at the absoluteness with which you stated your position. If you were describing your own methods, they're fine, but to make a blanket statement about having an accuracy less than a certain percentage is, in my opinion, insensitive and not particularly helpful to others who may be struggling, or just going along at 80% with no worries. Hopefully we all have the same goal, and the difference between 80 and 90% retention along that road need not invite blanket statements about whether one has truly "finished" the book.

Billyclyde, I was simply responding to the original question about what I found acceptable when doing my reviews. I think 90% is a good target, but if someone else were to argue for 100%, it wouldn't offend me.
Anyway, it is my opinion that if your goal is to learn all the kanji, then the most efficient way to get there is to aim for at least 90% whilst going through the book.
As for my comments about what constitutes truly finishing RTK, I was only pointing out, a little mischieviously I'll admit, that the closer to perfection one gets, the better. Put it this way, if you were running a marathon, but called for a taxi with several kilometers to go, could you really say you finished? By that formulation hardly anyone would ever actually complete the book, but I say aim as high as you can!

I think the recall rate is not really that important while you are doing RTK. If you follow the Leitner system, then you will see the failed cards all the time and you'll eventually get to remember them. The system adapts to your recall rates.

What's true is that if you have low recall rates then it will take longer to complete the book as you will be doing more reviews. But there's not much you can about that. Your recall rate reflects your ability at learning kanjis and that's something that will vary from person to person. It's inevitable that different people will have different recall rates. You can try to improve your story-making abilities and become better, but again not everybody will reach the same level. So, it's somewhat absurd to try to set a target number. The only thing you can do is accept your current recall rate and keep on working.

The recall rate that matters is when you've reached the end of the book. How good should it be? Considering that the 2042 kanjis are all important if you want to read Japanese, then a high recall rate would be necessary. But I think that depends on your level of Japanese. If you're still at a low level, then I think you could definitely content yourself with a lower recall rate and move on to other parts of the language. If you don't know some kanjis that you need, then you'll get to review them as you learn vocab.

That approach is somewhat sacrilegious to the Heisig method, and it's definitely cutting corners, but I think that it will work on a practical basis. You just have to keep on reviewing to eventually reach a higher recall rate. I think that an 80% rate would be somewhat of a minimum. That's 408 unknown kanjis. Below that, you would probably have too many holes in your knowledge to neatly fill them up afterwards.

As a final objective, 95% would probably be acceptable depending on the time it takes you to get there. That's still around 100 unknown kanjis, but again if you see them often you will end up remembering them. I think it's important to keep in mind that the time spent on Heisig is time not spent on learning something else. There's no point in being too much of a perfectionist

Just to clarify my original question, I'm thinking about the percentage of kanji you remember correctly while reviewing and studying RTK. Of course, eventually, one wants to remember all of them, but while learning and reviewing, it can't be expected you'll score 100 per cent every time. It just doesn't work that way. On the other hand, if you're only recalling half of the kanji you studied yesterday, I would definitely say you're going too fast, or not using Heisig's system to its potential. How fast you finish the book isn't the question here. Or how many kanji you know. It's figuring out an optimal percentage, or recall rate, let's say 90 per cent. If you remember more than that, you might think, gee, this is going well, I can speed up. If you score well below it, you think, whoa, that was terrible, I need to slow down and catch up a bit.
From what I've read here so far, it seems many people set 90 per cent as a benchmark from which to judge how well they're learning, and if they should slow down or can speed up.

I'd just like to stress that daily review pass rate doesn't equal long term retention rate. If you do 100 reviews a day and fail 20% that is just 20 kanji, or 1% of the whole book. Its not such a big deal.

Your long term retention rate will be higher than the daily review pass rate. Its the nature of spaced repetition; you will review new and problem kanji a lot more than the ones that you know well.

Actually, its probably worth talking a little about spaced repetition. We all should know by now that memorising lots of stuff is not just about working yourself harder, but also efficient reviewing. Its the job of the SRS to schedule your reviews at the correct time to achieve your desired retention rate. On more sophysticated SRS programs like Supermemo, you can set whatever desired retention you like. If you want 95% it will make it happen... by asking you to review much more often than if you pick 90%. (ReTK website uses a simple Leitner system that should work fine for most. Some users may be able to boost their memory by using a more adaptive SRS, like Mnemosyne, Anki or SuperMemo).

So here's my opinion. I think if you want the SRS to work well for you, you should follow what its telling you. If you have any failed or expired cards in any stack, its telling you to review them, so do it. This will optimize your memorisation. If you just rush to finish the book and leave cards to pile up, you're wasting some of your memory efficiency. The longer you wait, the more the trace of your memory will disappear and it'll become harder to learn.

timcampbell Wrote:Just to clarify my original question, I'm thinking about the percentage of kanji you remember correctly while reviewing and studying RTK. Of course, eventually, one wants to remember all of them, but while learning and reviewing, it can't be expected you'll score 100 per cent every time. It just doesn't work that way. On the other hand, if you're only recalling half of the kanji you studied yesterday, I would definitely say you're going too fast, or not using Heisig's system to its potential. How fast you finish the book isn't the question here. Or how many kanji you know. It's figuring out an optimal percentage, or recall rate, let's say 90 per cent. If you remember more than that, you might think, gee, this is going well, I can speed up. If you score well below it, you think, whoa, that was terrible, I need to slow down and catch up a bit.
From what I've read here so far, it seems many people set 90 per cent as a benchmark from which to judge how well they're learning, and if they should slow down or can speed up.

The thing is, everyone has different 'acceptance' requirements for when they consider something learned and ready for review or not. Do you review your kanji 2, minutes after learning, or 1 hour, or 24 hours?

Is a kanji considered learned right after you finished writing your story 1st time? Is it learned after a few hours? Is it learned after you've had at least 1 successful recall after 24 hours? Is it learned after its in the 4th stack? I'd say the latter, but some might have different ideas. Your percentage will vary based on when you measure. I personally expect a high percentage like 90%+ in the long term, but more like 80%+ in the short term.

You might want to use a long term benchmark, or just let the workload of the reviews guide you.

The further along I get into RTK the worse my L0-L2 rates get. There have been cards that I've gotten wrong 4-5 times before I got it right!

I can't let failed kanji stop me from adding new cards though. Otherwise, for me at least, it would take me forever to finish RTK1. I'd be stuck trying to get my recall rate up, while simultaneously dealing with expired cards (which happen in huge numbers lately). I don't think I'd have the energy or motivation to add any new ones.

So, even though my recall rates are steadily getting worse (with new kanji), I'm just pushing along. What you will find, regardless of your percentages, is that the stacks to the right will just keep getting taller Those stacks are the ones where I think your recall rate is important.

I think eventually, when we're finished with RTK1, we will all have excellent recall rates. I mean, think about it, no not as much new kanji, mostly reviews for a while. But for those of us still working with the book, I think it would be best to just keep pushing along and not focusing so much your percentages.

Now that I think about it, are you guys all finished? Because it sounds like you have amazing recall rates: 85%, 90%! Wow, I wish!

To me the Heisig approach (ie taking more time to learn the kanji in the first place) is best done by a student who can do his approach full time. But for those who can only spend a couple hours a day, I think it is better to rely on the Leitner system.
(IMO) to use the SRS to its full potential, you shouldn't worry about recall rates until you finish the book. Then take the time to focus on kanji that are causing trouble--then take the time to boost your recall rates by focusing on kanji that need to be focused on.

Quote:Is a kanji considered learned right after you finished writing your story 1st time? Is it learned after a few hours? Is it learned after you've had at least 1 successful recall after 24 hours? Is it learned after its in the 4th stack? I'd say the latter, but some might have different ideas.

I concur. It's an error to consider kanjis in the first or second stack to be learned. You might have tested these kanjis right after having memorizing them. The real test comes when you have let a few days go by. Also, if you rely on the first stack as a gauge of your performance, it can lead you to focus on the wrong things. It's easy to learn droves of kanjis and review them right after to keep your first stack green. I would advise to focus on the third stack and to consider that you have learned X number of kanjis only when they hit that stack. It's good for motivation too. If you focus on the first stack, you'll probably get depressed if you fail too many after a first review. I like to think of the first and second stacks as kanjis "on probation".

Quote:I can't let failed kanji stop me from adding new cards though. Otherwise, for me at least, it would take me forever to finish RTK1. I'd be stuck trying to get my recall rate up, while simultaneously dealing with expired cards (which happen in huge numbers lately). I don't think I'd have the energy or motivation to add any new ones.

I think this kind of situation is why Heisig is so much easier when you can devote sufficient time to it. If you keep pushing through too fast, then it will only get harder and you will be slowed down. On the other hand, I understand that your motivation will suffer if you don't keep learning new kanjis. That's why I think it's better to focus on the third stack. You should get your motivation from getting kanjis up to there and not focus on the first two. Every time you get a kanji to the third stack, you have learned a new kanji. Think of expired cards not as a chore, but as an opportunity to get those kanjis up to that magical third stack.

Quote:Just to clarify my original question, I'm thinking about the percentage of kanji you remember correctly while reviewing and studying RTK.

I was personally getting around 85% and up. That was when I was finishing the book and working at least 4 hours on it everyday. I had time not to let the expired cards pile up for more than 24 hours. If you let them sit there for a few days, than your recall rate will obviously be lower. As Vosmiura explained very well, you can have the recall rate you want depending on the frequency of reviews. And for a given frequency of reviews, your recall rate will depend upon your personal ability. My advice is to stick to your SRS system. Don't let expired cards pile up and don't worry too much about the recall rate. Of course, if it's something like 50% and you're following what the SRS is telling you, then perhaps you're doing something wrong.

Its usually about 80% or so. Well, it used to be. But about two weeks or so again, Ive begun trying something different.

See, Ive never been one for seeing through ones 'minds eye' or visualizing stuff in my head or anything like that, so Ive decided to stick to something a little closer to repetition. Basically I threw all I the Kanji I had gone through into the 5th box (though I didnt say Yes to whether or not I knew it if I really didnt)
Then I basically burned through about 100-150 Kanji a day in the 5th box. If I didnt know it, I would say no which of course would put it in the Failed box. Once I got about 10 in there, I would immediately go to them and put them in the 3rd box. Then at night I put them in the 4th, and in the morning the 5th.
At first I got up to 10 wrong in about 60 Kanji or so. Lately, however, Ive been able to go up to the 80s and 90s without getting a single one wrong, lately Ill just quit whenever I get tired rather than wait for 10. I think it's helping, since a combination of stories and repetition is more my thing than just pure story. The stories definitely help though, as I tried just pure 'write a new Kanji 10x times' before and it didnt work. I may not be able to see stuff in my head anywhere near as clearly as Heigsig, but he definitely helped me see Kanji in a logical way rather than just a mess of squiggles.