Personally, if I think two books are similar in difficulty, but one has furigana, I rate that one as easier. But even though I don’t know all the kanji yet, it’s easily possible for a book with full furigana to be more difficult than one without for me if it has a wider array of words, more difficult/specialized vocabulary and grammar, or uses a lot of non-Tokyo dialect. (This is especially true IME for manga, because they vary so much in how much dialogue there is.) Those are things that I weigh more heavily than the kanji usage when considering difficulty ratings, because those matter more in how hard something is for me to read.
3 Likes
I can’t do anything about it. The site favors comparisons between books that are at a similar level and I read so many books at the 30-35 level that every time I get 6 comparisons that are literally about the same in terms of difficulty 
4 Likes
I have a bell curve idea of difficulty where 30 is kind of standard, not too broad a swath of vocab, straight forward writing style, etc and most books fall somewhere from 25-35 with the outside ends being filled with children’s books + graded readers or literary (“college level”) writing + classic lit.
I’m sure some will disagree with that but it’s been my experience so not shocked it’s also been yours.
3 Likes
Nah, that seems pretty accurate for what I’ve read at least. I’ve read books/series at the following levels:
- 25 (children’s book)
- 25 (children’s book)
- 27 (novel)
- 27 (novel)
- 27 (children’s book)
- 28 (novel / short story collection)
- 28 (novel)
- 29 (LN)
- 30 (novel)
- 31 (LN)
- 32 (novel)
- 32 (novel)
- 34 (LN)
- 35 (LN)
- 35 (children’s book, but from like the 1930s)
Looking at the stuff I’ve read in the last year for reasonably accurate comparison, the level 35 LN and children’s book (魔法少女育成計画 and 銀河鉄道の夜) were an order of magnitude more difficult than the level 31 LN (本好きの下剋上), which is an order of magnitude more difficult than the level 27 novel (かがみの孤城).
3 Likes
Thank you for confirming, I was getting afraid I poisoned the database because of the curse of knowledge 
It might still be the case 
but at least the odds are lower than what I suddenly feared.
1 Like
Such descriptions being fuzzy is probably unavoidable. There is no clear line between beginner territory and intermediate territory, for example. But perhaps this is even for the best.
Who says they’re mutually exclusive? 
If people are fine with the extra bloat – or perhaps a toggle option for JLPT, CEFR, beginner-intermediate-advanced to choose between – then you could very well keep both of them.
This is mind-boggling to me. Seems very much skewed towards the lower levels. I agree that N1 should probably encroach into C1.
You could use the levels for different aspects, e.g. B2 reading comprehension. People have different strengths and weaknesses, so all aspects of one’s language proficiency aren’t necessarily going to be at the same level. Quite on the contrary, I’d say they are more likely not to be.
This is what I wanted to get across initially. The only reason I advocated for CEFR is that ‘B2’ is easier to comprehend at a glance, and takes up less space, than ‘Upper intermediate’.
There are certainly pros and cons for using either system. I am generally of the opinion that exam levels shouldn’t have anything to do on a platform where people want to learn a language, not study for an exam, but in this case I also see the benefit of using JLPT’s rough breakdown of expected kanji knowledge and such, like @basilsauce talked about.
Other than that, this thread is making me question my sense of reality, lol. I got a 134/180 on a recent mock exam for the N2 conducted by my school – which, granted, is a decent chunk over passing marks – but I’m also reading a novel and am very comfortable with it.
Speaking of, rumour has it that こう, the keeper of jpdb.io, which has all manner of statistics for the works in its database, is likely willing to collaborate with Natively and cross-link databases. jpdb–Natively crossover when???
3 Likes
Also N2 being in part A2 feels strange. Since the equivalence was made by the jlcat (the same people providing the CEFR-based test), they have a bit of a conflict of interest here so I think it’s fair to say that the somewhat bumped down the JLPT.
Yes, that’s a decent chunk. In comparison, I passed the N2 with a 0 point (!) margin back in the days, and reading (it’s hard to believe at this point) was my weakest score at the time. Meanwhile, I was comfortable reading novels when I first attempted the N1 (and failed it by 3 points), so I was still N2. There’s a huge gap between barely passing N2 and barely failing/passing the N1.
3 Likes
Lots of great discussion going on here, wish I’d come back sooner 
JLPT LEVELS
Per the JLPT discussion, I’ll just say that it’s a very valid suggestion to move the ‘JLPT’ recommendations to the grading page and out of the filters / grading popups as a way to de-emphasize them and encourage relative thinking. The JLPT recommendations may be the most frequent critique I get. However, that complaint almost always comes from users who already know the level system and don’t need it, which makes things hard to judge 
.
As a contrast, I constantly see people in forums requesting resources for different JLPT levels and then receiving links to Natively for appropriate searches. And, like I said before, I do think they’re pretty good recommendations for those levels, at least about as good (and way more comprehensive) than most blog posts. Yes, the leveling system doesn’t evaluate JLPT guidelines and therefore I can’t objectively point to why this absolutely is a good JLPT recommendation, but I don’t think that necessarily prohibits the fact that it is still a pretty good recommendation.
If we strongly feel that this isn’t the case and that the JLPT recommendations are substantially worse than other blog posts out there recommending specific material for JLPT learners, than I think this is a moot point and I should get rid of them.
If we simply think that we should push users away from that thinking which is why we should de-emphasize them, then I think that still warrants discussion and perhaps it’s right. But I am a little wary because it is a good potential selling point (SEO, linking) and growth is a focus right now. So if it’s 50/50, I’d probably lean towards keeping, if that makes sense.
As you can probably tell, I do lean towards keeping them for the time being. I think it’s good to contrast it with JPDB which doesn’t try to do this mapping and I think it’s much less intuitive for new users.
SWITCHING TO CEFR
As for switching to CEFR, I do agree it’s a more comprehensive grading system and very well might work better for audiovisual. Audiovisual may again prompt this discussion! I do like the beginner-intermediate-advanced approach and CEFR as a proxy for them. But, like I said before, JLPT has a lot going for it just with the institutional knowledge about those levels in the Japanese community.
2 Likes
Mmm… let’s just try to get more grades, seems easier and more robust 
But yes, there is that potential.
Potentially! If people just get annoyed by seeing a certain level scheme and just want to view different ones, could work.
2 Likes
As an American who has never learned a language other than Japanese, I have no what the CEFR levels are. At least I know what the JLPT levels are.
Is there any reason not to have each language on Natively map to a system used for that language? Since these are just rough guidelines anyway (the Natively levels are what truly matter) it seems weird to pick a “one size fits all” approach.
5 Likes
I’m personally fine with the JLPT levels even though they’re imprecise. Really any subjective review by masses of amateurs (as opposed to say professional teachers) will be, but imprecise is better than nothing! I think probably the lower levels (0-20?) are pretty robustly mapped to actual JLPT content because of graded readers and beginners noping out harder at unknown grammar and vocab, but beyond that point it probably gets fuzzy in relation to the test and I think that’s ok.
As for the CEFR levels - I’m familiar with them and prefer them for myself (since I think of them via the Can Do statements), but am apathetic about them being used on the site. I do think most Japanese learners won’t understand them and will simply ask for JLPT leveling. A toggle could work, but then you’d get the 
fun of extended discussions on the mapping of say, level 30 / N2 / B1 or whatever. 
2 Likes
Actually, one thing I just realized earlier in this thread is that the JLPT levels are not linear, but exponential in terms of difficulty. N1 is twice as much knowledge as N2, which in turns is twice as much as N3, and so on. However, the scaling with levels is linear (except for N5 which covers 12 levels, everything corresponds to 6 level). I guess that’s one reason why there’s a discrepancy between the two types of ranking? Maybe give 3 levels to N4 (13~15), 6 to N3 (16~21), 12 to N2 (22~31), and everything else to N1? Obviously, those are random values, there’s probably better adjustment available, but it would take into account the difference in ability between when I was a low N2 and a high N2.
2 Likes
That’s indeed a considerable difference. I guess this, and all the other viewpoints provided in the thread as a whole, just goes to show just how different the make-up of one’s language ability can be from person to person. To paraphrase what someone mentioned earlier in the thread, the point at which you can start reading novels is going to be much earlier if your learning has been mainly geared towards reading.
1 Like
In the same vein: It feels weird to me that some books have a “fixed” grading with only one person ever having graded it.
I graded ユートピア | L38 ten times (with the extra gradings I think?) and now its rating is fixed, or for example 三四郎 | L40 has 20 gradings by one user only. Maybe implementing a 2 or 3 user minimum would work against such a bias?
For me, at least, I feel uncomfortable with the fact that my ratings on ユートピア made it a fixed “level 38” book. Of course subsequent ratings influence the rating as well, but somehow to me, even just seeing the tag without the “??” makes me think: “Oh, looks like it’s been graded by several people to be at about that level.” maybe that’s a me problem though 
Since I also commented on only starting to read books when I was well past N2: For me it wasn’t so much that I wouldn’t have been able to read novels/manga earlier, but I somehow just never got around to it, never really tried picking up a book. (Maybe partly due to the fact that I wouldn’t even have known where to start, which book to pick, level wise, etc.) My frustration tolerance is also pretty low, so I think this approach was right for me.
Which led to me being able to read my first book pretty comfortably, and I didn’t really have a “dissecting and looking up tons of words to make sense of this” phase. I highly respect anyone starting to read at N4 or even N3! The earlier you start, the more ahead you are on the learning curve as well, since reading books also takes practice.
3 Likes
I also sometimes feel like at least two users should grade a work for it to get a grade without a “?”. But brandon might not like it if suddenly half the database has an uncertain grading.
I personally feel good about the grades I gave for manga. But not so much about the novels. Both because I’m not quite used to novels yet and because natively asks me to compete manga and novels to each other, which I just can’t do very well. Reading novels just feels harder to me than reading manga in general.
@brandon is it possible to delete and/or redo ones gradings?
2 Likes
Yes, you just need to go to https://learnnatively.com/grading-list/… should consolidate that with profile gradings page…
… I know people continue to talk about this, but i do want more grades in the db first. I also think it’s less of a deal than people think, but maybe i’m the one wrong there.
Most people say よつばと! (series) | L17 is quite comfortable at the end of Genki II, and yotsubato fluctuates between L17-19, so the N4 gradings seem right to me. While I agree that there is an exponential difference between N3, N2 & N1, remember that ELO is also not linear, but a bell curve… so I don’t necessarily think that the levels have to be exponential in size?
I also don’t think N1 should be ‘everything else’ as its only B2! N1+ should have the really hard stuff like it has now, with lots of classics. My impression was that the ranges were in an Ok place now, but we can relitigate if people want… perhaps in a different thread.
2 Likes
Not sure what “profile gradings page” refers to, but I kept looking for that functionality on the page of the book I want to change my gradings for.
4 Likes
That’s fair enough! 
Same, right now changing gradings is really annoying. Mostly I look around the book’s page and compare my gradings with others, and then realize “okay maybe I should change this rating”. But instead of being able to change the grading right then and there, I have to go look for the page with my gradings, remember which comparison I want to edit, find that book, etc. etc.
But I think you’re already aware of people wanting easier access to this function ^^
4 Likes
I gave those levels completely at random as an example
Well, it depends on where you put the center of that bell curve, compared to the JLPT levels and how the two scaling compare to each other. It might not be that we need to double every time, but it still might be that a constant number of levels per N is not the best match. I don’t know.
Actually, now that I look at the actual books available at the N3 level, I feel like I could have managed 時をかける少女 back when I passed the N2 with a 0 point margin (which we might as well call a “solid N3”) and that book is at the upper range of N3, so the levels may work better than I expected? Also, 獣の奏者 1 was pain at the time, and it is at the upper range of N2, but I would have managed it at the time I was nearly N1 as well…
But then the original comment of @basilsauce still holds: I would have been destroyed by また、同じ夢を見ていた, despite its easier level, due to the lack of furigana. However, that book has 150+ ratings across 27 users, so it doesn’t feel like a random score. I guess there’s not much one can do about that…
Sure! I forgot about it when I was writing, but I would have gone with the same logic… not that it really matters at this point.
1 Like
Is it possible to allow users to filter by furigana? We’ve had some talk about whether people are factoring that into their gradings or not. But as @onigiristudies said in the grading discussion thread, it would be really nice if users have the ability to check a box to return furigana results only. As much as I advocate for making the site accessible to traditional students, it’s also definitely true that there are a lot of simply written books which are solidly in the 23 to 30 range in every single aspect except for the lack of furigana.
I’d also like to add that there are a variety of reasons why someone might want to be able to browse a range of full-furigana books, even at higher (30+) levels, beyond just “don’t know the kanji yet because they did school order.”
-
They used RTK or know Chinese
-
They’re re-starting Japanese after a break and still understand grammar but forgot a lot of vocab
-
They still need to look up a lot of words
-
Related — they want to look up a lot of words (level 44 夜は短し歩けよ乙女 has a Tsubasa Bunko version, not yet listed on Natively）
5 Likes
