My main problem with Wanikani

Hi there

I’m a happy owner of lifetime access. After struggling for years with the RTK book and with Anki decks, I was won over by the quality of WK mnemonics and cross-references. I’ve been doing WK for exactly 1 year now and I just recently got to level 30, commonly thought of as “half the way,” although it will probably be 1/4 of the way for me, if I’m lucky (more on that later.)

One thing I’ve been doing all this time, because I was accustomed to it from all the years I spent on Anki, is back-to-back cards. I have configured my apps so that everytime I review a card I get asked 1. first the reading, 2. then the meaning for the same card. If I get both right, the card advances to the next stage, otherwise it recedes. This has always made sense to me: when I see a Japanese word I want to be able to read it out in Japanese, and only then I want to recall its meaning in another language.

I did try doing reviews with the default settings (independent reading and meaning cards) and it was extremely confusing: I was constantly typing the reading in the meaning box or vice versa, or thinking about the reading in order to recall the meaning (which is what I normally do) but not being checked on it therefore getting both wrong. A complete disaster.

Now that I’m at level 30 though, I’m starting to have a big problem: I keep finding Enlightened and Burned items that I completely forgot. Like, when I review the mnemonics I go: Uh? Did I ever study this in the first place!? I have no recollection of it what-so-ever.

On many kanji and words, it seems I’m able to complete all reviews and get them to Burned without actually committing them to long-term memory. I have no idea how I can do that. Or maybe they get there and then escape, I’m not sure.

The reason I wrote this post, and my main gripe with Wanikani, is that it does not have a button (nor an API) to “nudge down” a specific kanji whenever I realize I forgot it.

Let’s say both 辛 and 幸 are Enlightened or Burned, or 喉 and 候, or such pairs. I could go on forever, I have dozens of kanji pairs that I keep mistaking for one another, way past their Burned status. And let’s say I come across a word (either in a WK word review, or in a book I’m reading) and realize that I either forgot that kanji completely, or I learned a new kanji very similar to it and it got “replaced” in my memory (which happens very often.)

What I would like to do is navigate to the kanji page and hit a button that says “Oops… I forgot it!” and it will nudge the kanji down 1 or 2 steps, say from Enlightened/Burned to Guru or Master.

Yes, there is a dodgy Resurrect option, but it has a few problems: 1. it will place it back into Apprentice I, which is ridiculous; 2. it doesn’t work while the kanji is Enlightened, which is where 80% of my instances of Oops happen; 3. it’s not a real API, so apps have trouble placing the relevant button in the kanji info page.

So this is the main problem I’m having with Wanikani. I thought it was a missing interface feature (the Forgot button) but then it got me thinking: is nobody else plagued with forgotten items that were previously Enlightened/Burned? Am I doing something wrong in my study?

Should I ditch the back-to-back cards? Is that feature playing tricks with my memory? Or is it because English is my second language? Maybe I’m linking Japanese kanji and words to a different part of my brain, compared to people who study it with respect to their first language. But I’ve been using English for most of my adult life, so that would be weird.

Any advice will be very welcome.

I might try to do a few levels with the independent card system (if I can get used to it) and see if I get different results long-term.


With the enlightened I don’t see much trouble, because if you are forgetting them, when they come for the burn review you’ll fail them and you will, have to review them quite a few time more, so they might stick better that time.

What seems to happen to me is that I can mostly get reviews right in the first tries, so I most of the time take them to master or enlightened with no trouble, but when they come for either enlighten or burn is like my memory is completely blank, as you say, did I even see this kanji before? Maybe the system is more thought for people that will forget them in the early stages and those of us whose brain might not forget them fast have more trouble later on. I’d say most of the items I’ve burned (which are up to level 10 approximately) are mostly ok in my memory, it is true that I might have some confusion between items but that’s not that much about memory, because I do actually remember them, but I confuse the overall form. I think that’s something that will only be solved with more exposure.


For similar kanji, these two userscripts have been a big help to me:

When I mistake a kanji for another, I can just open the info and see where I went wrong and how to better differentiate between them.

Also, for vocabulary and readings there’s this script that tries to guess where the confusion arose:


You write that you have a problem with WK, but your problem is with you doing 1x1; Kouichi purposely made vanilla WK to separate the meaning and reading so that you review them separately to better reinforce them, so you are not piggybacking the one off the other. Drop the 1x1 and you retention will improve dramatically.


I’ve had the very same problem with forgetting some of the Burned items when I’ve seen them somewhere else and then I resurrected them to Apprentice, however, I kept thinking that Guru or Master might have been better…

What might mitigate this problem for you could be the Self-Study script w/ the Additional filters (include Burned items in Self Study reviews):

I’m really just repeating myself here, though since I’ve made a nearly similar post here.

Some time has elapsed since then, and I’ve decided to entirely reset my progress from level 45, since life got kind of really busy (my son was born and I kept forgetting everything I’ve learned up until then).

Now I would give a different kind of advice, just 3 things:

  • Don’t worry about forgetting, even burned items, it’s just part of the learning process.
  • Start reading/keep reading native material, that way even though you come across something that you feel you’ve already learned, it’ll just get cemented further in your memory.
  • Lastly, I think the interleaved cards of WK are a great help. If you’re having a hard time dealing with the reviews, that just means it works! The more you have to struggle find the solutions, the stronger those memories are going to be linked in your mind (I mean, that’s what the creators of the site say too, and I agree with them)

@Rowena That’s pretty speculative. I’m not convinced using 1x1 mode makes a significant difference one way or the other, and even if it sometimes does you certainly can’t claim changing this will solve OP’s problems.

Regardless of 1x1 mode, it’s pretty common to get kanji mixed up as you learn more. @Gymgirl claims this is fine even for Enlightened items since you’ll just fail them the next time you see them, but I think this argument is flawed. First, if the Enlightened review is in a few days, you might actually Burn it because you just relearned the word/kanji a few days prior. It didn’t really test that you remembered it for months like it’s supposed to. Second, the Enlightened review could be in 3.5 months, and waiting that long for more review of an item you’re struggling with is silly.

Being able to manually demote any item to any lower SRS rank would be very helpful. Speaking from experience, I use a feature like this on Kitsun fairly often. I see a word when reading that I already learned but forgot. I demote the word from wherever it is (sometimes Burned, but often not) down to around Guru rank. I see no point in demoting it all the way to Apprentice 1. I frankly don’t understand why WaniKani hasn’t added this feature yet, even if just through the API.


Burned items aside, since you can only revive them, for enlightened items, you can use the “undo/typo” script to purposefully fail an item during reviews if you don’t think you know it well enough. Since the item doesn’t actually pass/fail in the WK system until you move onto the next item, this script gives you the chance to look at the mnemonic and think abut how well you know it, and then you can personally determine if you want it to pass up or fail down.

However, as @seanblue said, a “down nudge” feature would be helpful


Yeah it doesnt.

What does this even mean? There is literally no such thing as piggybacking srs items, whatever the hell that is supposed to be lol.

He is just speculating that is the cause, and I can assure you its not. I did scattered reviews for a long time and back to back reviews for a long time and saw no drop in retention. Plenty of other people share a similar experience.

Yeah, no.


People like to make fun of Anki but the fact that you can indicate which items you’re struggling with (as well as some automatically applied multiplier depending on how often you get them wrong) is actually a feature. So is being able to change the review intervals.

As for the “similar kanji” issue, I have the same problem, I often find my memory being “overwritten” (or at least confused) by similar Kanji. What’s more, this doesn’t really matter in the Apprentice/Guru stages because I can just recall “it has to be this Kanji because the other one I’ve already burned/is in enlightened” etc. On the other hand, if you get two kanji mixed up, you’ll probably still be able to read, as context would dictate which one it is (or the vocab, e.g. 幸せ). I’m thinking that maybe after WK I will specifically make (or find) Anki decks that test on similar Kanji. Or, alternatively, I try to learn handwriting them.


I would recommend Anki if flexibility is what’s the priority here.

On a more personal note, I’m not sure whether long-term repeating English glosses is worth it. Probably long-time learners would be able to tell better :slight_smile: . For me it feels sufficient to know what the word means and how to use it, so the “feeling” behind it and not the translation to another language. That’s why I don’t have 2 sets of cards in Anki for each word, but only 1 set (JP + example sentences → EN with multiple meanings or additional explanations)


I love Anki and for vocab it’s perfect, but I do believe that WK has some things to add specifically for Kanji.

That said, I’m currently using this approach:
I do WK and I also make a bidirectional Anki deck for my Genki vocab. When I add an item to Anki that I don’t know all of the Kanji for, I add it in hiragana. If it comes up in a review and I have learned the Kanji on WK, I then change the card to the Kanji version. I often also add WK vocab to Anki too, unless I really know it very well.
That way, both tools reinforce themselves. It’s working really great for me, I would say that I have almost 100% recognition on my Anki vocab (in both directions).

Edit: I also set up my Anki cards so that I have to input the Japanese (not the English, though). That way, if I’ve already Kanji-fied my card, I’m still training the reading by having to input it through the IME.

1 Like

Hi, sorry to read about your troubles. What I do not see in your posting is what it is you’re doing with what you have learned so far. As in: are you reading manga, books, playing games or anything where what you have learned so far could come up? Or have you set your goal to first do WK, then grammar, then start using it in real life? Because, personally, I sincerely doubt it’s possible to retain all that information you’ve learned here so far by simply doing reviews (within WK or elsewhere). Actively using Japanese should greatly help you improve (and expand) your retention. The rest (WK, anki, etc.) are just training wheels.


Is this your experience? If so, I’m glad it worked for you, and I’m interested to know when you noticed this approach worked better.

However, I always do my best to recall meaning, reading and kanji at the same time, and I’m of the opposite opinion: I think it’s harmful to separate them, because in real life, you’re going to need to see a word, pronounce it, and understand it, or hear it and understand it (possibly while visualising it or feeling some sort of visual trigger activating); or you’re going to feel like saying something and you’re going to ask yourself what you know that is close to what you mean. I think studies of speed-readers have even shown that those who claim to understand without mentally saying words end up doing it to some extent, because understanding is almost impossible otherwise. You need all three elements to be linked to each other, at least at the word level. (It may not be as important at the kanji level.) That brings me to my next point:

With all due respect, ‘piggybacking’ is a natural part of human memory. It’s called ‘associative memory’. We all have an easier time remembering things based on something we already know, and it’s one of the most effective ways of memorising something. Memory athletes, for instance, use familiar things like the names and faces of famous people to help them memorise packs of cards at contests. WaniKani relies on this as well: that’s the reason why recurrent characters are used for reading mnemonics, because so many kanji readings are the same. I’ll grant that it’s preferable to be able to remember something without having to spend a ton of time on some other trigger first, but that’s why memorable mnemonics are important.

Perhaps more importantly, trying to completely separate the things one needs to learn can cause other problems. I’ll quote this post, which explains it much better than me:

(I’d say that the full post is worth a glance, at the very least.)

Granted, this doesn’t mean that it’s bad to test yourself separately on meanings and readings, but this seems to suggest that learning them together is more helpful, so testing yourself on meaning and reading pairs also makes sense.

PS: Just as a final thought: when I was learning Chinese as a child, we definitely were taught meanings and readings together. Our vocabulary lists were always structured the same way: characters for a word, readings, meanings, examples. Forgetting one or the other (or both) is perfectly natural, but I can say for sure that we were tested on meanings and readings both separately (through spelling tests/dictation and comprehension questions) and together (through sentence-making spelling tests where we had to write words down and then make a sentence that showed what the word meant). Neither form of review was significantly more effective in my experience, just more or less difficult. What mattered most was how well we had mastered the words beforehand, and that depended on how we had learnt them.


I just want to be clear here, the official WK stance is called interleaving: Randomized Review Order | WaniKani Knowledge


That FAQ entry touches on two things, a) mixing up different types of items during a review (Kanji, radicals, vocab) and b) asking for the meaning first sometimes and for the reading first some other times.
I don’t believe it applies to artificially separating the meaning and reading reviews really.

Additionally, when they write “it has been shown”, I would love for them to cite specific studies so that one can verify what exactly has been shown. (Or alternatively, they should just write “we believe”).


You’ve basically said what I was going to add, except to point out that:

Character amnesia is real and it affects native Japanese speakers as well as non-native speakers. Actually using your newly gained kanji knowledge is the best way to retain it.

And… my new word of the day (from the above Wikipedia article) is ワープロ馬鹿:


I guess that makes sense for reviews, especially if those things appear within the same review session. What I’m saying is that it’s probably a good idea to apply a unified logic to learning. WK doesn’t say that’s bad though, so I’m not going to add any criticism here. Nonetheless, just as a personal note, I can more or less guarantee that this is what would happen if someone asked me to use such a review system, unless I was really out of time/trying to see how fast I can finish my reviews:

  1. Answer the question asked
  2. Get myself to recall the element not requested for fun – if I haven’t already – just to make the review harder for myself and to provide an extra challenge

You could call it nothing but a silly, cherished belief, that’s for sure, but to me, that’s how it goes: I learnt these things together, and I will need to use them together, so I’ll recall them together as much as I can.

I think I’d start making changes here. If you’re wondering how, once again, I recommend reading the post I linked to earlier for the basic idea:

While I’m very much in favour of linking readings and meanings together, there’s one huge problem with needing to use the reading to recall the meaning: Japanese contains too many words (and kanji) that are pronounced exactly the same way, even if you account for pitch accents, which most of us don’t study very often to begin with. Case in point: I had a Japanese test less than a week ago. We were given a list of words, almost all of them in written only in hiragana, and we were told in translate them into French. (I’m doing my degree in France.) There was one word I couldn’t translate until an hour after the test started (I did the other questions first): しゅうしょくする. If I had only studied Japanese with this teacher, I would have had no problem. But because I already speak Chinese and I’ve noticed several rules that let me convert pronunciations in Mandarin into Japanese pronunciations, I was stuck. ‘Which しゅうしょく? What is it used for? There are so many words that seem like possibilities.’ I only came up with an answer after remembering that we did a chapter on life stories and milestones, which meant it was probably 就職=‘to gain employment, to get a job’. Point is, readings are far from unique, so how are you going to deal with overlapping, seeing as you can’t prevent it? Similarly, you can’t just link a reading to a kanji if your objective is to recall its meaning, because some kanji sound similar precisely because they look similar. (They’re pictophonetic/phonosemantic kanji (形声文字) i.e. kanji with a meaning component and a sound component.)

What I’d suggest is that you try to link both meaning and reading to the kanji at the same time, by having them cooperate in some way. Ideally, your memory should be something like a 3D matrix or an XYZ-coordinate system: knowing any two of the three elements (kanji, meaning, reading) should help you find the third. I’ll work with your examples:

Fun fact: in Mandarin, these two actually sound almost the same (one just has an extra G that gives it a more nasal sound, and they have different tones), so they’re even easier to mix up. I definitely mixed them up as a child. How can you remember them? A few ideas:

  1. Try learning words you can actually use/you’ve come across before that use them: 幸福(こうふく)= ‘happiness’ does not look like 辛苦(しんく)= ‘hardship’. 幸せ(しあわせ = ‘happiness’)does not look (or sound) like 辛い(つらい = ‘difficult, painful’ or からい = ‘spicy’).
  2. Associate them with real context that makes them memorable: you’ll probably find 辛 or 辛口(からくち)written in violent, rough calligraphy on spicy instant noodle packets, like this:
    See if you can find a Japanese ramen place near you that offers such ramen if you really want to strengthen the experience. It helps with remembering other meanings too: I love spicy food, sure, but I’m not going to deny that it can be ‘painful’ to eat. Hahaha.
    For 幸, I can’t think of an example right now, but I can guarantee you there’s a manga somewhere out there with 幸せ written in bright colours with lots of cheerful 〜〜〜〜 lines, possibly floating above characters’ heads as they relax in an onsen (or as they eat a delicious hotpot).
  3. Play spot the difference: this phrase is favourite on the Japanese Internet when someone asks what the difference between two expressions is: ‘They’re different where they’re different, and the same where they’re the same.’ This may not be etymologically correct at all, but when you’re looking at two similar kanji, pretend that the reason they’re different is the parts where they don’t match, and then make up an explanation. For example, here, we’ve got ‘、’ versus ‘十’. You can pretend the dot (’、’) is a drop of acid, and that it burns, so it’s ‘painful’; or that it’s a chili seed, so it’s really spicy; or that it’s a pebble that got thrown onto someone’s head. (Pretend that’s the rest of the character.) For ‘十’, it’s easy: it looks like a plus (+), so it’s positive and ‘happy’!

You can apply the same strategies here. I’m not going to go into detail though; I’ll just tell you how I differentiate them:

  • 喉’s got a mouth (口) in it, so it’s obviously linked to the mouth somehow. The ‘throat’ is linked to the mouth alright. What about the reading? Well, it’s のど, and I don’t know about you, but both の and ど require me to round my mouth nicely, and if I freeze right after I say ど, I look like a goldfish. (I’m exaggerating a little, obviously.) What I imagine is that as I say ど – because my mouth is open and everything seems to frame my ‘throat’ nicely – I get one of my fingers and poke the back of my ‘throat’ like an idiot. (Ok, the truth is that I just concentrate my attention on the back of my ‘throat’ as I say ど, but it’s essentially the same thing: the pointing finger of my mental hand brings my ‘throat’ to my attention.)
  • 候 tends to have meanings that are related to time or moments in time – waiting, serving, checking on someone/something, the weather at a particular point in time or in a particular season (a bit of a stretch, but you can see how these things are all somehow related to time)… What’s it got that the other kanji doesn’t have? A line: l. What now? You can pretend that’s a line on a clock (and I vaguely remember a teacher telling me that it was added as a symbol of something like that). You could also imagine it’s someone attentively waiting, as one would when checking on someone at home or in hospital, or when one is in line, like when one is a candidate(候補者 – こうほしゃ)for something.

There you go. I hope some of those ideas help.

It would be nice for them to do so, yes, but for what it’s worth, I remember being given a list of study tips by a neuropsychologist when I was younger and recovering from a head injury, and there was reference inside it to a study on interleaving. Everything was cited for people who were interested. There’s also a course on Coursera called Learning How to Learn that recommends interleaving, again with research cited somewhere in the course, probably at the end of that chapter. Studies on this exist, and while it would have been better if WK cited its sources, in this case, I can assure you it’s well researched, at the least if the medical community is to retain its credibility. WK isn’t pulling this out of a hat. I’m just too tired to bring up a study myself right now. If you want a rough idea of the research methodology though, typically, two groups of students were taught something and asked to review it in a certain way: one group working on one topic at a time, and the other mixing topics. Ultimately, the group that mixed topics did better. I believe the most commonly cited study discusses students doing maths problems. I don’t know if others exist though.

Agreed. Practice is key. There are Chinese characters whose readings I can’t remember for certain (or which I can’t write anymore) because I just haven’t needed them in ages. I’ve been speaking Chinese since I was a baby so… yeah, it can happen to anyone.


I’ve read what I think are probably the same studies referred to. Just my opinion, but I think it’s a difference of interpretation of the study results. The research talks about studying related topics, such as math subjects that would typically appear as back-to-back topics, and interleaving brings the previous topic(s) back while quizzing for the current topic. The theory is that interleaving helps the brain see relationships between related topics that would otherwise go unnoticed if you didn’t quiz them together.

So, again in my opinion, the interleaving (a) brings together topics that would otherwise be separated by days or weeks, not minutes as within a kanji review session, and (b) is intended to bring together things that would otherwise be separated, not separate things that would otherwise be together such as reading and meaning.

Based on prior testamonials from people who have tried 1x1 mode, I think the main down side is that separating the reading and meaning naturally brings up the item twice in a review, which is extra practice, whereas back-to-back is essentially equivalent to the item being brought up once. But there is an easy remedy for that, too: Rather than doing reading-then-meaning back to back, you can simply recite both in your head (or, better yet, out loud) no matter if it’s a reading question or a meaning question. From the experiences I’ve heard, that method seems to overcome any degradation suffered from 1x1 mode., and typically significantly improves review accuracy. (For me, 1x1 alone was sufficient.)


辛 and 幸

Here’s another one, from 1:23 to 1:33:


I do think WK could offer more flexibility on the system to let you review something more (demote items, freeze leeches, etc). Scripts aside, the system is rather rigid which makes sense, but there can be some leeway implemented.

I also agree with interacting with Japanese more. Our brains have a harder time remembering things we don’t use, especially the amount of items you have to juggle in WK. Seeing stuff in context, writing it, hearing it and using it will always help retention more than srs I feel