Back

JLPT QUESTION about the scoring system

#1
Hello everyone! So lately I've been wondering about how the scoring works in JLPT's exams... I mean, for example, their website states that to pass the JLPTN5's listening part, you have to get 19 points out of 60 (or 33%). If the listening section has 28 questions, it means you get to pass just by answering 9 questions? Or does it require to pass a certain amount of questions of each mondai too? 


Thanks in advance for anyone who's too kind to answer my question!  Blush
Edited: 2017-07-01, 6:23 am by ファブリス
Reply
#2
The scoring system for JLPT is complicated. Basically every question is automatically weighted to be worth more or less depending on how many people answered it correctly (easy questions most people got right being worth less than hard questions that not many people got right). Then they apply the 19/60 checks and the overall passmark check. So you don't need a particular score in each little 'mondai' subsection of the test. But on the other hand it's not as simple as "ok if you got 9/28 right" because if you answered only the easiest 9 they might be worth less than a full point each...
Reply
#3
(2017-07-01, 4:17 am)pm215 Wrote: The scoring system for JLPT is complicated. Basically every question is automatically weighted to be worth more or less depending on how many people answered it correctly (easy questions most people got right being worth less than hard questions that not many people got right). Then they apply the 19/60 checks and the overall passmark check. So you don't need a particular score in each little 'mondai' subsection of the test. But on the other hand it's not as simple as "ok if you got 9/28 right" because if you answered only the easiest 9 they might be worth less than a full point each...

I passed the N5 last year.  I read the JLPT detailed document about how the exam is scored, and from what I could gather, they take the raw marks and "adjust" them so that it looks like the exams are at a consistent level of difficulty from one year to the next.

So for example, if they screw up and make the questions for the N5 too hard compared with prior years, they adjust the marks upward.  So I suppose there are people who would have failed based on their raw score but squeaked through with a marginal pass after the marks are "adjusted."  On the other hand if they accidentally made the exam too easy they would adjust the marks downward.

I could be misinterpreting what they're saying, though.
Reply
EPIC SALE: Get 30% OFF Premium & Premium PLUS! (Sept 18 - 29)
JapanesePod101
#4
I think the process is not as human-mediated as your description makes it sound. As far as I can tell they have a clever algorithm such that they can throw everybody's answers in at the top and it automatically determines which questions were hard and need to be heavily weighted, and which were too easy and need to be lightly weighted, and then everybody's resulting score drops out the bottom (and as you say, in a way that scores work out consistent from year to year). The wikipedia article on Item Response Theory is a bit confusing for me though.
Reply
#5
(2017-07-01, 1:11 pm)pm215 Wrote: I think the process is not as human-mediated as your description makes it sound. As far as I can tell they have a clever algorithm such that they can throw everybody's answers in at the top and it automatically determines which questions were hard and need to be heavily weighted, and which were too easy and need to be lightly weighted, and then everybody's resulting score drops out the bottom (and as you say, in a way that scores work out consistent from year to year). The wikipedia article on Item Response Theory is a bit confusing for me though.

I believe the document about scoring said something to the effect that "no matter how hard the committee [of people who decide the questions] tries, sometimes the level of difficulty doesn't match the prior years."

So the algorithm is meant to correct for the human error (it was humans who came up with the questions) which resulted in setting the exam too difficult or too easy.
Reply