I don’t know that it’s actually disallowed… It just wouldn’t mean much since a native would be expected to show far higher proficiency than what is tested.
Something like the Nihongo Kentei would be more appropriate.
5 Likes
The FAQ page say JLPT is for non-native speakers but how they actually enforce/define it, I have no idea. I do see Japanese children in US take the test. Since there is a grading scale with a ~30% pass rate, these are good questions. Kind of handy to have your mom and/or dad be your own personal tutor, right?
2 Likes
It would make sense if they were to disallow it, as scores are normalised
2 Likes
Kind of, but not in the way you’re thinking, I don’t think. It’s not adjusted after the fact; other students’ scores can’t affect yours.
What they do is take all the questions as written, then have experts decide how ‘hard’ they should be for someone at the level being tested. So maybe there’s an N2-level question on the N4 exam. That’s ok! (A bit demoralizing, but…) They just weight that one so that missing it doesn’t count against you as much as getting it right counts FOR you.
It’s graded holistically, in that every possible combination of correct and incorrect answers is evaluated for proficiency. Someone could miss three questions on three different concepts X, Y and Z, but get others on those same concepts right. Someone else might miss three questions all about X. Same raw score, different result. If X is critical, the second person might fail while the first person is ok. That’s why they can’t say a question is worth N-number of points.
Now, although it’s not graded on a curve retroactively, that doesn’t mean a native-speaker acing the test wouldn’t cause a problem. It might screw up the pass/fail statistics enough to undermine confidence in the test as a diagnostic tool. It could artificially influence the scoring experts to make the next year’s test harder. But it wouldn’t affect any other test takers that same cycle.
Edit: Sorry, I’d like to cite my sources, but I’d never be able to retrace how I pieced all that together back when I was insanely interested. I sometimes do written testing at work, and I thought it might work for us that way. Turned out, it’s way too much work up front for me to be able to get routine tests out in a timely manner.
6 Likes
Not sure I understand how this is defined. When I look holistic grading, it’s essentially a rubric for compositional grading (?) and not scaled scoring for IRT to keep test quality standards consistent. Do you have source to help me understand?
My understanding we are still getting grouped into and high or low performers based on answer patterns in the spectrum of test takers (or at least according to this source). So wouldn’t abnormal high performers skew the analytics if current test takers are defining the spectrum…possibly kicking out a borderline passer?
1 Like
Not all test takers, as in real test takers on test day. Sample test-takers, before the exam is given. Experts who aren’t really answering for themselves, but judging what a N5-level should be able to answer. This is the standard by which all the actual test-takers are measured. Theoretically, every test taker could pass. (But that would be statistically unlikely)
That’s what keeps the test this year from having a different “pass” standard than last year or the year before, which is a problem with curve grading systems that depend on student results. That way the scores are comparable to previous years even if the average competence of the tested group is different, or if the test questions were objectively harder or easier.
Edit: but I’m going to add the same disclaimer as the guy on that website you linked. I researched it on my own, and this is the understanding I came away with. But my understanding is nowhere directly confirmed by the owners of the JLPT, just indirectly supported by some of the things they do say. One important takeaway I remember (but now don’t remember where I found it, naturally 
) is that it absolutely does not depend on the other test takers in your cohort.
Edit 2: I started with this pdf from the JLPT web site, then followed the references at the end. The pdf seems to support, but not outright confirm, my theory.
2 Likes
The pdf mentions IRT. IRT is a common method used to “scale” tests such as the GRE and GMAT.
Typically, IRT does not use pre-exam sample test takers, nor evaluation by experts.
Rather, IRT is calculated from the real test scores obtained from the test day. That is, IRT computes the difficulty of each question, based on the “raw” scores from test takers taking the same exam. (*)
There are different IRT methods, but conceptually here’s how IRT works:
For any given exam, we have test takers with resulting raw scores ranging from low to high. A question which everyone tends to answer correctly might be weighed as an “easy” question (low parameter weight). A question where those with higher raw scores tend to answer correctly more often than those with lower raw scores might be weighed as a “hard” question (high parameter weight). A question where only the highest scoring test takers consistently answer correctly might be weighed “super hard”.
(In reality the parameter weights are continuous, e.g., from -2 to 2 where -2 is the easiest and 2 is the hardest, following a certain distribution curve.)
Now that we know the difficulty parameter for each question, we can produce a scaled score that can be compared against other exams taken in different periods (e.g., last year’s exam). In this way the exam difficulty can be kept consistent over time.
(*) To add: a sufficiently large sample of raw results might be used for calculations, instead of computing the scale using all results.
2 Likes
You may be right, but I object to that assumption. The difficulty relative to the ability of the students who took it would be consistent, but there would be no reason to believe this year’s exam was the same difficulty as last year’s, because the tested group was different.
Maybe the error is negligible enough with such a large sample.
Like I said, I could be wrong, but something I read made me SURE it was a control group of educators, not real students that the question difficulty ratings came from. Now I’m doubting myself. 
Generally, one of the goals of IRT is exactly so exam difficulty from year to year can be compared (and adjustments to scores can be made if desired).
Indeed one of the reasons IRT was introduced into the JLPT was so that the new “N” levels could be compared to the “old” JLPT levels (with old exam IRT weights computed “after the fact”).
E.g., the reason MEXT can state that the new JLPT N1 is equivalent to the “old” JLPT 1, or that JLPT N3 is “between” the old JLPT 2 and 3 levels, is presumably because they’ve calculated the IRT weights and made adjustments as necessary.
Here’s the quotation from one of the linked references:
The passing standard for the new test is based on statistical analysis and designed to match that of the old test. Examinees able to pass Levels 1, 2, 3 and 4 in the old test are judged to have the Japanese-language proficiency to be able to pass N1, N2, N4 and N5 in the new test. As for N3, there was no corresponding level in the old test, and passing scores for N3 are designed to fall between the passing standards of the old test’s Levels 2 and 3, based on statistical analysis of Japanese-language proficiency levels for passing these levels.
2 Likes
Agreed, all sources are showing scaled IRT is an effort to keep score difficulty consistent. I never heard of pre-exams, thought it was all post exam evals that determines final scaling. From my first link, they said the raw uncut vs scaled IRT is within ~0.95, so the scaling is not terribly extreme. Given there are no prerequisite or requirements for enrollment, this is probably reasonable as the spectrum of skill level may not be consistent between the years. If we get examinees with a perfect score, it’s not a clear indicator of consistent test difficulty or whether someone picked levels well within their ability. However if we hypothetically enroll 50,000 certified N1 passers in the N5 test, will that mistaken N2 level question by JLPT now get flagged as the accuracy spectrum changes?
The site does say the intent of the test is for non-native speakers. However, I think we’re in a gray area. I don’t think this is an enforced rule because the gray area can be quite difficult to account for every situation.
The kids are US citizens who never attended school in Japan and never lived full time in Japan. (Actually they did attend the last two weeks of a public elementary school one year during summer break just for the experience and the school welcomed them eagerly. Story in itself.) I think in the questions during sign up, they asked about some background and I did answer that Japanese is spoken at home. I answered all survey questions fully and accurately. They were happy to take my $60 times 3.
Plus it is on the radar screen for my daughter to attend college in Japan as a foreigner next year. It’s a fairly unlikely path, but we want to have JLPT credentials prepared for her should we pursue this. I do not believe that her Japanese abilities will be on par with the average entering college freshman and college texts in Japanese would be a huge stretch for her. So, though she has spoken Japanese her entire speaking life, I would not say she is on-par with a native 18 year old Japanese citizen. She would probably take courses in English while sprinkling in a few Japanese ones to see how things went.
10 Likes
Not when they are teenagers… trust me!
8 Likes
I finished 日本語の森’s new GO KAKU series for N1 and really enjoyed it. However I just feel really burnt out of studying.
I’m finding that my usual methods of studying-but-not-really-studying not to be working this time around. I’m still reading, playing video games, watching movies/Let’s Plays and stuff but I realize that all the content is a little too easy. I rarely ever run into any N1 grammar points, which makes sense since they’re all rather formal/old fashioned, but it’s just incredibly frustrating.
I don’t want have to always read the news/look at grammar books to keep being exposed to them. I’m also getting burnt out from all the SRS for Vocab.
And we only have a month and a half left. 
How is everyone else doing?
2 Likes
Panic is slowly setting in.
They withdrew the money for the test. No turning back.
But oh man, I am sooo behind. Especially on vocabulary and Kanji (i was supposed to have reached level 20ish by now instead of 17). I severely underestimated the vocabulary section. Grammar should be fine, as long as I don’t skip any more days and cram until due date.
Listening section however… all the practice so far seems to be going great, so at least there’s that. LOL
save me pls 
8 Likes
I haven’t taken any practice tests, so I have no idea what to expect for N3. On grammar I’ve gone through half of Tobira and a third of Bunpro N3. Unlikely I’ll finish either of them and I definitely won’t have time to review with Shin Kanzen Master. Kanji will obviously be fine and vocab should be fine based on the vocab used in the Tobira readings. And I never practice listening. I’m definitely most concerned about listening, followed by grammar.
1 Like
The Nihongo no Mori N3 series are really helping me with listening since they teach N3 in Japanese. It’s a bit of a struggle keeping up sometimes, but I can feel progress after nearly completing the vocab section. It’s very interesting to hear him re-define vocab in relatively simple Japanese.
I own a copy of Tobira, but it really feels even drier than Genki. These texts feel like a cruel form of torture to go through.
2 Likes
I am working through Tobira, and indeed the texts… I don’t really want to read them anymore.
I have been focusing on the grammar part as I’m feeling behind, so I’m currently rushing through that (at lesson 9 atm) and in the final week, i’ll read through all the texts and review the parts I don’t understand. It’s not ideal, but since time is running out, this is what I’ll go for.
Also Nihongo no Mori is definitely recommended. Japanese all the way. I use it to reinforce what I know.
3 Likes
I’m not a fan of Tobira. It is really dry, to the point that I only read the quick grammar explanations and occasionally the reading for a chapter when I feel like it.
Maybe I should check out Nihongo no Mori too…
3 Likes
@Saida have you received your certificate from Leiden yet? Some have gotten it yesterday already, but I haven’t gotten anything yet, so I was wondering whether I’m the only one
I haven’t checked my mailbox since おととい, but others in my class haven’t gotten it yet, either.