I have to wonder about my long term retention rate. I don't usually do "review all expired." Instead I do each pile separately (failed cards, 1 review, 2 reviews, 3 reviews, and then 4 reviews piles). Today I got 100% on 1 review, 100% on 2 reviews, 92% on 3 reviews, and then 71% on 4 reviews. (The reason the first piles were so highly scored, I think, is because I've stopped at 1750 cards, and so those cards are mostly old cards that I had to review again). I'm not sure how many of the "4 reviews" were actually 4, 5, or 6 reviews. (I don't have anything in box 8 yet.)

Is 71% bad? It seems pretty low and so it seems like maybe the intervals are too long for me. Or maybe the way I think of my stories are too short-term oriented. Some I don't think I will ever forget, like "brandish" and picturing the guy from Gladiator riding around the stadium on a Chariot brandishing a big "we're #1" type foam Finger. A lot of the ones I'm remembering I think I'm just remembering what primitives go in them, because I've seen them so many times. But that makes me wonder if I will just forget everything once they hit the long delays after 5, 6, 7 reviews.

Or is it normal for 25-30% to be forgotten once they hit the long term intervals?

I just checked and the most fails I have on a card is 5. Most of those I remember having a lot of trouble with but I'm pretty sure I won't forget them anytime soon because I've had to review them so many times until I got it right. That is somewhat comforting.

I was curious what everyone's pie charts look like. Do you eventually get almost all of them into "pile 8"?

Here's mine:



Maybe I should just hurry up and finish (and keep reviewing after that of course) but starting learning more about the kanji (KO2001 and other reading) to help really seal them in. It seems like some of the ones I've seen in use get easier to remember because I have more different mental connections to that kanji.

Forgetting the occasional kanji is no big deal, though you may want to take a closer look at any individual kanji you fail too often.

My chart isn't really accurate anymore, since I migrated to Anki about a year ago, so I won't bother showing it. Generally speaking, though, most of my failed cards seemed to be in the middle piles (3, 4, occasionally 5). Any cards that made it to pile 6+ pretty much stayed memorized.

(Edit: Though I'll admit, this chart thing made me a bit curious. I may go through my expired pile when I have some time to spare.)

I think I'm in a similar place, although I've actually finished RTK1. If I'm being truthful, I would say I typically end up in the mid-80%s for mixed reviews of all boxes. So, I'd say we're doing about the same. Even though it pisses me off when I miss a few, I try not to worry too much. The SRS is doing its work, so ones I need to review, I review. One thing to keep in mind is how you're missing them. If it's a single stroke or a confused placement, or just a matter of not remembering which of Heisig's confusingly similar keywords goes with which kanji, then I don't worry. I might tweak the story a bit, but usually I just let it be. If I totally blank, and even seeing the kanji doesn't recall the story, I rethink my story.

71% may be on the low side, but the real problem with retention issues at this stage IMO is the possibility of rep buildup and burnout. If you aren't adding new cards particularly quickly (or at all, it sounds like), I don't think it'll be a problem.

~J

Let me see what I missed...

4 wrong/missing primitive
3 blank
1 wrong kanji/keyword

consent. I knew it was words + either young or juvenile. I was going to go with young at first but then went with juvenile.

consummate. For some reason I thought of the wrong connotation of the word. The adjective as in "complete in every detail" instead of the verb "to make complete (marital union) by sexual intercourse" so I blanked on what to put here at all.

accept. birdhouse + crotch. I put "every" instead of "crotch" because my story was a birdhouse brothel that accepted any crotch (which I remembered as "every").

pedestal. I mixed this up with podium. I always get those two confused even though I can write either one just fine.

split. I did bone + garment instead of bone + saber on the top and garment on the bottom.

holy. complete blank.

task. I was trying to piece this together from my vague story (in my mind it represented the story from the wizard of oz). I got halberd + broom + power instead of halberd + taskmaster + power.

un-. complete blank. To me it was "jail" and I forgot that as a kanji it's un- (although, this was more of a short term one compared to the other ones above)

I wouldn't worry too much about a low retention rate, as those Kanji you missed will come back sooner.

I find that with me it differs greatly one day or another. I don't have any time to add new Kanji at the moment (hopefully this week or next week I can start again, it's not a long-term stop), but I do my reviews every day, just to remember those that I've learnt.

Yesterday I had to review 27 and I failed 6. Today I had to do 33 and I failed only one. I've find it often differs (though 6 was reallllly bad). I can't tell you what causes it though...

71% doesn't sound too bad. 100%, 100% and 92% for the others is great.

The thing about doing Heisig reviews is that you never encounter them outside your Heisig reviews and studies. Whereas when you do vocab, what you put into your SRS regularly pops up in your other Japanese studies/experiences.

As long as you keep practicing your retention rate will slowly climb. Most of my misses seem to be misconceptions, of the sort that drivers99 seemed to describe (e.g. "young" vs. "juvenile"). I'm especially bad with the kanjis that seem to have the same meaning (rebuke, punish, chastise, etc).

We just have to keep ironing these out until they become reflexive, which is the whole point of studying a language after all, isn't it?

Don't sweat your retention on a day to day basis. Instead, if you can, go with your overall average retention. This is easy with Anki as it tracks that information, but will take a bit of work with RevTK's srs since you'd have to write down your daily results.