These are just manual mappings, so super easy to have language specific. I agree with Myria that people who aren’t experienced with CEFR (i.e. have studied romance languages) probably have little conception of what they mean. The government tests have specific standards and, in general, the specific language communities have an already developed sense of what their test numbers mean.
If it were a big issue to have different ones for each language, it might be a consideration… but not a big issue
Considering that the CEFR is generally used in Europe, it’s not too helpful when a majority of languages or the countries associated with them do have their own levels for their own proficiency exams. It’s more helpful in the longrun to keep the JLPT tags for the Japanese books because for those who aren’t European (or whose country doesn’t subscribed to that framework), the A1-C2 levels mean absolutely nothing. At least the JLPT is consistent no matter which country.
Like @Myria, I’d definitely find it confusing for it to be on there. It’s similar to when people ask me what my CEFR level is and I tell them that I’m not European, so I’m not studying with that frame/goal in mind. I’m studying with the proficiency exam involving my language in mind. Even for Korean, a lot of books will label it with what TOPIK level it will correspond to because it was created by Korea.
There seems to be misconceptions about what the CEFR actually is, so let me explain: the CEFR is simply a system used to divide language proficiency into levels. A common way of breaking down proficiency in any skill is beginner, intermediate, and advanced. The CEFR takes it a step further:
It has nothing to do with exams or European languages. It’s merely a description of one’s proficiency in a language, which is why it has such a broad area of application, unlike language-specific language exams. All it means to pass a certain level in a language exam, is that you passed that level in the language exam. It has surprisingly little to do with actual language ability. This also has other implications, such as language exam levels not being a very good indicator of the level of media, which further gives sort of a false promise that if you if you passed a given exam level, you should expect to handle media tagged with that level. This will inevitably lead to people going in with such an expectation, only to be overwhelmingly crushed by it.
Ultimately, I simply meant to make a suggestion that, if you are going to use a system for breaking down language ability into levels, why not just use arguably the most widely used system in the world, which is also language-independent? Easy to implement, and users only have to get used to one system once, assuming they are not already familiar with it. I obviously don’t know what it looks like from the implementer’s side of things, but I figured the input could be worth communicating, and then brandon can consider whether that has a place in his vision or not.
Yes! That’s a totally reasonable view point and I appreciate the feedback. I’ve thought about just doing the CEFR levels and perhaps may change my mind in the future after more languages are added - it’s a good suggestion.
I don’t disagree that language proficiency tests are not a great marker of proficiency (JLPT requires no speaking… really??), but I think they’re the only standard that most people in a language community know.
CEFR is a generalized framework with generalized descriptions for each of those levels… but those generalized descriptions are of course fuzzy too.
Nothing is really perfect here… many people have suggested I ditch the JLPT mappings entirely. But, don’t let ‘perfect’ be the enemy of ‘good’ as they say… and users, especially new ones, get a lot of value out of the mappings, flawed as they might be.
And I do think they’re pretty good suggestion as is.
N2 being (in part) A2 just sounds so funny
I guess another point against CEFR here is that one of the reasons the JLPT compares so badly is the missing speaking and writing components, which is not something you actually need for reading books.
I definitely agree that N1 is nowhere near C2, especially when taking production into account. But passing N1 does prove quite a high reading comprehension ability in my opinion.
But maybe you wouldn’t even have to directly map the N levels onto CEFR.
Another option would of course be to get rid of the „N“ labels when you hover over the book level, and only in the „About“ section go into more detail which level would correspond (roughly) to which JLPT level. In the same table, it would be quite easy to add a CEFR column as well without cluttering.
In that sense I think I actually agree with @Arctagon, just that maybe instead of using a framework not everyone might be familiar with directly in the tags, using descriptors like „upper intermediate“ would be my preferred approach (with more information in the About section).
This way you also remove the disappointment of „why is this N2 Level book so difficult even though I passed N2?“
Which is what B2 would be, by the way. At least, I did pass the N1 when I was definitely still (low) B2, which is why their chart makes sense to me. I still feel like someone who got the N1 with like 90% should be considered C1, though.
By the way, every book was too difficult for me after I passed N2 I sincerely think people are amazing for reading novels at that level.
Fair enough I only started reading my first book a few months before I took N1, so I can relate. Thinking back on it, my reading comprehension probably did not exceed B2 level either
I use the JLPT system to guide my gradings and reviews. I guess I never asked about it, but I’ve been assuming that ideally Natively will be useful to as many learners as possible, and that this includes people following a traditional curriculum at school.
Schools all over the world use textbooks designed to follow the JLPT system, and there are specific sets of kanji, vocabulary, and grammar patterns expected for each level. The JLPT organization no longer publishes official lists since the change to the N~ system, but aside from N3 which is new, the other levels still map fairly well to the lists from the old system. The people who’ve taken classes at language schools or university (or even followed a textbook course faithfully) are pretty consistent not only in how their abilities are balanced, but even to which specific kanji and grammar patterns they know.
Language schools prepare their students for the JLPT with these kinds of numbers
N5: 100 kanji
N4: 300 kanji
N3: 650 kanji
N2: 1,000 kanji
N1: 2,000 kanji
In other words, if a book has 1300 kanji and no furigana, traditionally taught students at N3 level can’t read it. For that matter, the N2 students might have a rough time too.
I, like @Naphthalene, found novels too difficult after passing N2. But I did it anyway, at a huge cost. Nobody was giving recommendations that were appropriate for my level (or gender), so I chose my own material and made some big mistakes. I read a couple of children’s books that were really hard, and then grownup books which were WAY too hard, pushing myself as much as I could. After a year of this, I was much better at reading and also so burnt out that I gave up on fiction for years. My Bookmeter account is 12 years old and there’s a 10 year gap in the middle. If it wasn’t for covid closing the borders so that I could no longer visit Japan and get practice/maintenance chances, I might never have picked up another book or manga in Japanese again. I wish something like Natively had existed so that my experiences could have been easier.
I can really relate to that. A lot of things I have seen strongly recommended online turned out to be aimed at a different demographic than my own. That was fine at first, but it grew old quite fast.
That’s where I stopped at the time. I read the first volume of the children edition of 獣の奏者 (which took me 6 months at the time) and tried to read 楽しいムーミン一家 but gave up. I only really started reading books when I got near N1. Manga, on the other hand, where fine as early as N4 (well, that was series dependent though)
Generally speaking, I feel like the relative ranking on Natively (level) is more important than the N something equivalent.
I think that’s why a lot of people learn kanji at a faster pace than a traditional course or textbook would assign. I read my first book (with much difficulty) when my grammar was between N5 and N4 level, but I knew something like 1000-1200 kanji (at least). It was a kids book (魔女の宅急便), so the kanji wasn’t much of an issue for me. At this point I’ve read about 30 books / light novels that use “normal” amounts of kanji for adult readers, but I’ve never studied N2 grammar, let alone N1 grammar. If I had to guess, I think I’d pass the grammar and reading sections of the N2 test, but I’d probably fail those sections (or at least struggle/guess a lot) in N1.
This is one reason I don’t think the JLPT levels are good indicators for reading level. It’s easy to bias your learning regimen to target reading, while neglecting what it would take to pass a higher level of the JLPT (even not accounting for listening). Like @Naphthalene I think the relative Natively levels are far more important than how they map to JLPT level.
I won’t chime in to the JLPT discussion at hand yet (it’s late here!)… but I wanted to just say I’m starting to do an outreach push which will continue for a little while, perhaps a month or two.
I tweeted and made a reddit post, both with relatively good fanfare. The most exciting aspect is all the connections with Japanese book review bloggers I’m beginning to form and hopefully foster more growth.
I also want to give a shoutout to @cat 's about section getting a little press:
No, it doesn’t necessarily have to do with European languages, but the framework was created in Europe and is used across Europe. Though many other countries are adopting the framework, it’s not common everywhere. It was developed in Europe (Common European Framework of Reference for Languages), so though it is considered to be international, it doesn’t mean that every country uses it. For example, America typically uses the ACTFL system. That’s not gonna mean much to other people if they aren’t within an area that’s using it. I’m aware of the CEFR framework because I tutor kids from European countries who are learning English (and using that framework), but it serves little use for me where I live because it’s not our standard here. If I’m tutoring someone from a country that doesn’t use the CEFR framework, the safest thing to fall back on is the TOEIC because it’s well-known.
What I was saying is that when you’re purchasing materials to study Japanese that were made in/from Japan and not necessarily by contracted foreigners, they are generally labeled with the JLPT levels because, as far as I know, the CEFR exams aren’t offered in Japan. While it is true that the JLPT doesn’t measure speaking (which, in some instances, is nonsense in it’s own right), Natively isn’t about speaking either. It’s about reading and, soon, listening comprehension, so I don’t think that using the CEFR levels instead of JLPT will make a huge impact. I do think it would be odd to see a textbook that has the words JLPT N3 in the title/on the cover labeled with the CEFR framework.
I definitely think that while everything may not be perfect, the JLPT mapping is a great spot. Especially because the CEFR framework isn’t used everywhere. So if somebody who’s from a country that doesn’t use the CEFR framework (or has never heard of it) were to come onto the site, it might be a loss for them. It would begin to get aggravating after a while to have to continually convert the scale to match another one when there’s already one set by a specific standard, where the tests are offered in multiple countries.
I don’t think books necessarily need to map exactly to JLPT levels, but if we assume that everyone is following an AJATT-style model of going hard on kanji early on, then it risks excluding a substantial population of students. I’m not saying that we all need to make kanji level a dealbreaker. However, I strongly believe that books for below-N1 (or “high intermediate”) readers should have information in the reviews about the kanji level, and existence (or not) of furigana. And it’s definitely not the case that we can’t find books intended for native speakers which accidentally overlap well with JLPT kanji/grammar levels.
魔女の宅急便 is one of the books that matches surprisingly well with both the online idea of “beginner” and N4 grammar/kanji. The kanji level is very limited. You didn’t need all of those 1,200 kanji to read it. The kanji printed without furigana are only the ones which a 2nd or 3rd grader would know, and that has a lot of overlap with the kanji expected for N4. ふしぎ駄菓子屋 銭天堂 is another series which is quite accessible to traditional students. また同じ夢を見ていた is not. It might have a similar vocabulary and grammar level, but the kanji is N2+ unless somebody publishes a furigana version. I think books like that should have a heads-up in the reviews. (I left one)
Another reason why I try to consider the JLPT standards is because I think it helps me maintain consistency in my comparisons even as my own Japanese level improves. There were several books that I read, or tried to read, in the past year which were much harder than I had expected based on the level that another user had rated it. (This isn’t just one person, there are multiple advanced readers who appear to be overly optimistic about what beginner to low-intermediate readers can handle) I started several books graded as similar to コンビニ人間 which actually turned out to have over twice as many new words per page and a whole pile of new grammar patterns to look up. I chose these books at コンビニ人間 level because that’s what I felt comfortable reading at that moment, and wasn’t expecting to be feel overwhelmed instead. It was incredibly frustrating and discouraging, and I had a hard time sorting how much of my experience was because the book was an unsuitable recommendation, and how much was because I’m a failure who can’t read at my level. I know those advanced readers only wanted to be helpful, and were only hoping to expand the options that their juniors could choose from. But those were some painful experiences for me, and it taught me that it’s really hard to gauge the difficulty of material that’s far from your own level.
As my reading is improving, I can tell I’m increasingly at risk of doing that myself now. Just because the book I’m reading now feels easier than my memory of what コンビニ人間 felt like a year ago doesn’t mean it really is. The JLPT standards (especially the grammar) help give me some obvious reference points to keep an eye out for.
(edit: this isn’t really a reply to Naphthalene, I was just writing, but I can’t seem to edit the “reply” part)
Yes, that is surely true. That’s just where I happened to be when I read it.
I’m sure people appreciate that information being included, but personally I focus more on the content than the difficulty. I do sometimes mention something being “easy”, but I never get into specifics. And partly that’s because it all depends on how a person studied.
For example, in my review of かがみの孤城 I wrote “In terms of difficulty, it’s among the easiest Japanese books that I’ve read”. This is true based on how I judge difficulty, which is primarily sentence structure, grammar usage, and word usage, etc. But it’s mostly just based on how it felt to read and not a particularly deep analysis. Since kanji is one of my stronger abilities with Japanese, that doesn’t come into play as often. But for someone following a “traditional” learning path, かがみの孤城 would probably still be difficult at N2 and someone at N3 would stand no chance. (A full furigana version did recently come out that may allow those “traditional path” N3/N2 learners to enjoy the book.)
Barring any kind of difficulty breakdown, is there anything to be done about this? After all, people are just comparing the difficulty of the books and then it’s up to Natively to decide what to do with that information. Perhaps at some point @brandon should add some kind of “bias” to the algorithm that accounts for users that are unusually high or low in their ratings. (Looking at you @Naphthalene, for rating almost everything as being “about the same”. ) Probably not feasible until more people join and rate books, and if there are enough ratings it may not help much anyway.
This is definitely a risk. I feel like my analysis of かがみの孤城 being an easy book is accurate, but it’s possible that I’m simply not accounting for my improved abilities over the years. (Though to be fair, I skip all gradings from the first handful of books I read to at least partly avoid this.) That said, because I haven’t followed a traditional learning path since my first year of learning Japanese (going on 6.5 years now), it’s really difficult for me to analyze difficulty in terms of specific things like kanji, vocab, grammar, etc. So all I can really do is base it on gut feel. Again, hopefully this gets worked out over time as more people join Natively and rate the same books. At least this should work out for more popular books/series.
@brandon Another idea here would be to get counts for things like total character count and unique kanji count, in addition to the page count you already get. That way people can decide whether to read a book not solely based on the Natively difficulty level. Some websites also provide unique word count, though those can be error prone due to misparses. I doubt the Amazon API you’re using has this information, but maybe you can find a way to get it at some point.
This is something that I also have to keep in mind for myself! I noticed it when I submitting books to be added to the website that I considered to not be too difficult for my level, but then I realized it’s because those topics interest me, so the vocabulary wasn’t an issue. I enjoy yakuza series and royalty series, so I noticed I was grading them lower than they may have been when I submitted them. I read ヲタクに恋は難しい 1 | L28 with my bookclub and found it to be ridiculously difficult because I don’t know gaming terms even in English.
Based off of what you said, I’ll try to make sure that when I’m reviewing things, I’ll include if it was something I’ve already had interest in, or if it was an entirely new topic.
It also feels like to me that the difficultly levels generally assume people can read kanji without furigana. I find it hard to believe so many people can do this, but the ratings are coming from people rating some books without furigana easier than other books or manga with full furigana so maybe this is the case???
It a difficult subject because once you can read many kanji, then it is actually easier to read books with a lot of kanji rather than kids bunko books with full furigana, which also tend to that a lot of onomatopoeia and hiragana verbs. So I would actually rate books like these as easier book.
On the other hand, if I find myself looking up a lot of characters and readings I rate something more difficult.
For example 2 books that I consider that the kanji choice, compounds, and furigana choice make it more difficult compared to the sentence length, grammar, vocab, etc. are…
That might be the WaniKani bias. I think a significant number of early Natively users came from WaniKani, which means they are bound to know more kanji than average.
Yeah I personally never think about furigana when it comes to rating the difficulty of a book. I generally rate based on how complex I find the grammar to be and what the vocabulary in the book is like.
I would personally have N1 encroach into C1
I sincerely think people are amazing for reading novels at that level.
