I’m done with Wanikani, but I thought I’d post some feedback here for anyone who might read it, in case it’s of use to a dev, or more likely, a fellow frustrated learner.
I have reached a saturation point with my frustration with this app twice now, at level 24 and then at level 27. Both times I’ve become so viscerally angry while doing WK reviews that I’ve reset to level 20 just to make my review queue go away. I don’t think I’ll recover this time.
Levels 20-29 is called ‘death’ in the Wanikani categorisation - and it feels like a stark jump up from Levels 1-19. It’s full of really, really irritating kanji - things like 突, ‘stab’, pronounced ‘totsu’, which for some reason shares nothing at all except its shape with 空, ‘sky’, ‘kuu’, and 究, ‘research’, ‘kyuu’. This ten-level block seems packed with kanji that are complex, not terribly easy to distinguish from other kanji, and with wildly different, non-phonetic readings. It’s also packed with gems like 触る and 触れる, which have different, unrelated readings but almost identical meanings (‘to touch’), and the latter looks like potential or passive to boot (which it is not). Quite frankly, I hate this.
This is clearly not entirely Wanikani’s fault - Japanese is hopelessly complex, even for native speakers, and notoriously very, very difficult for native English speakers to learn. The rage I feel when doing WK reviews is not rage at the app, but at myself; it’s at my perceived lack of progress. My deep resentment of the kanji might in fact mean Japanese is not for me. But I don’t experience the same bursts of rage when using my other learning apps, which suggests they make me feel like I’m actually learning in a way WK doesn’t. After three-odd years, I feel like Wanikani has made a range of choices in app and curriculum design that are more of an impediment than an accelerant in my learning.
My personal ‘dealbreakers’ for Wanikani
-
Lack of integration of context sentences: Context sentences in Wanikani are buried down the bottom of the page, and play no part in learning or reviewing vocab. They are also presented in a way that doesn’t call attention to the word being learned, and are often too long and complex for the level of the learner. Worse, they use multiple different translations, and the sentences are often chosen for their ‘funny’ translations rather than their utility to the learner. I found in the early days I just had no hope of understanding them, so I gave up. Three years in, revisiting them for this post, I can actually read them now, and some of them seem kind of useful - but they are still buried at the bottom of the page, and are often too long for what they are trying to achieve. Regardless, for me the ship had already sailed - I tried a few times int he first few weeks and then wrote them off permanently. I know Wanikani is not a vocab-learning app but all language learning is done through context. I think having a very simple, single-clause or at most two-clause sentence for each word that clearly demonstrates its most simple usage would be a huge benefit for learning vocab. Better yet, instead of making me enter the reading on an error-prone keyboard with no ‘whoops’ button (#2 below), let me just read the kanji/vocab in context as a review - I’ve come around to the input hypothesis that comprehensible input is way more important than output, and at the intermediate plateau, comprehensible input is really hard to come by.
-
Poor error management: It’s been talked about ad nauseam on this forum, but the app needs a ‘whoops’ button. It also needs to engage in a positive, encouraging way with your ‘near misses’, for example where you type ‘せつ’ instead of ‘さつ’. At the moment it treats a serious but wrong attempt the same way as it treats me inputting ‘お’ in frustration because it’s the closest key to the enter key and I want to skip to the screen that tells me the answer. On that note, why do I have to manually open the info screen after an error, and then individually open reading and meaning tabs to review the item? There’s not even an option to fix this, it just assumes you enjoy beating your head against a brick wall. For me, I’d rather see near-misses bounced back with targeted guidance on what I’m getting wrong (maybe highlight the ‘せ’ so I know it’s meant to be ‘さ’?), or if I put a valid but wrong reading for the kanji in, a hint that it’s a different reading. I’d also like a setting that shows me the meaning and reading instantly on a wrong answer so I can refresh myself on it.
-
Inflexible SRS and poor leech management: If a Kanji or vocab word is still bouncing between Guru and Apprentice after ten levels, let me unlearn it and come back later, or just abandon learning it until I encounter it in the wild. It’s clearly not sticking, and just showing it to me on repeat for months doesn’t feel great. In fact, it feels like I’m not learning anything - which, it turns out, is pedagogically and psychologically a sure-fire way to not learn anything. I’ve also never got used to, and always hated, the insurmountable-feeling and disheartening flood of end-of-level vocab. I want an ‘unlearn’ button that makes the item just go away until I’m ready to come back and try learning it fresh later, and an ‘I know this’ button that keeps basic stuff they added later out of my already ridiculously large review queue. And more than anything else, I want a mode where you learn one kanji, its reading, and all its associated vocab, so you can avoid the enormous, morale-sapping flood of lessons at the end of every level.
-
Flippant attitude to difficulty: I don’t need pass-agg notes that I ‘should be able to read this already’ on my vocab cards, especially not in a language where there are multiple irregular readings for an enormous number of Kanji. I need the bloody pronunciation of the word. See also - overly complex context sentences (#1), minimal support for ‘near misses’ (#3), assumption of insane grind as a precondition of learning (#5). The app is built around ‘number of reviews/lessons goes down’ as a philosophy of language learning - in other words, it’s made a fundamentally qualitative exercise of learning through context into a quantitative, gamified process of rote memorisation. It sets up an expectation you’ll learn all the kanji in a year (delete this slogan NOW), offers a mathy path to achieving it (so many lessons, so many reviews, such and such an accuracy %), and yet, the app doesn’t really do anything to help number go down, and seems to gloat at you when number in fact goes up.
-
Tendency to overwhelm time-poor learners: I read one ‘level 60 in a year’ post that said this required 400 reviews a day - and knowing what I know about WK, that number assumes you’re getting 80%+ right. This program needs to make space for the 50% and 60% accuracy crowd, It needs to be honest that it’s likely a 5-year undertaking for mere mortals. It needs to be upfront that you’ll have enormous review queue and an exhausting lesson queue at almost all times unless you optimise the hell out of it. OR, and this is my preferred method of dealing with this problem - it needs to eliminate the numbers from the game as much as possible, because there are people who find numbers incredibly demotivating. Don’t show total lessons in the queue, for example, or hide the panel of on-level kanji that lets me ‘game’ the level-up system. In Bunpro I never feel the need to grind out to N3, I just plod along doing my lessons and reviews. Bunpro doesn’t seem to want to rub my face in how slow I’m going, WK really does. I don’t know why, but it feels like a UX/UI problem, given the similarities in the apps.
-
Reliance on scripts to fix the programs’ most egregious flaws: Every problem above - or, almost every one, given there’s no fix to the rigid SRS - can be fixed by scripts. For some of the more esoteric stuff above, that’s fine. But for basic things like a ‘whoops’ button, kicking your users to a script database is pretty egregious. WK isn’t Anki - be flexible, or be prescriptive. Don’t try and be both at once. And if you are going to be prescriptive, be honest.
So that’s it - that’s my beef with WK. I think I’m done. I’ll likely burn what’s left and finish up at level 20, then I might try and get the Anki WK deck and pick up where I left off, trying some of the fixes I talk about above. I’m really annoyed right now, but despite my frustrations, Wanikani has worked really well for the kanji it has managed to teach me. Well enough I don’t regret the lifetime membership I bought only last year after paying a yearly fee for two years.
Obviously none of this means that WK is a bad program - it just means it’s a bad one for me. Luckily I’ve found one that’s excellent for me - Bunpro fixes most of the flaws of WK’s SRS, mainly by offering a really powerful set of tuning options and a much more flexible SRS. The only downside is it’s not so crash hot for Kanji, but then, it’s not meant to be. But I’ll figure something out. Best of luck to those of you struggling the way I have.
) will suggest you to start with some formal course like Bunpro (by internet) or Minna no nihongo (if you prefer books).
