The JLPT tests language knowledge, reading and listening, and nothing else. There is no speaking section and no writing section at any level, so a candidate can hold the band employers ask for and never once have been assessed on producing a sentence. That is not a flaw in the exam, which does what it says. It is a fact about what the certificate is evidence of.
Because the test defines the market, the schools, the textbooks and the apps all optimise for recognition, so a learner following the obvious route can spend three years becoming very good at understanding Japanese and remain unable to hold a conversation.
The examination, in six lines
- N5 to N1, with N1 the hardest, which is the reverse of the direction most numbering schemes run and catches people out.
- there is no speaking or writing test. The JLPT assesses recognition only, which is its central limitation and the reason a certificate and a conversation can diverge so widely.
- N2 is the usual work requirement for employment in Japan outside specialist roles, with N1 asked for in professional and academic settings.
- language knowledge, reading and listening are the three sections, with vocabulary and grammar folded into the first.
- it is held twice a year, in July and December, though not every level is offered at every overseas site.
- N1 does not mean fluent. It means a very large passive vocabulary and fast reading, and holders routinely report being unable to hold the conversation their certificate implies, which is the honest thing to say about it.
Passing is a total and a floor in every section
The structure is worth knowing before booking, because it catches people who prepared by the total. The paper is scored in sections and you have to clear both an overall pass mark and a minimum in each section separately, so a strong reading score does not rescue a weak listening one. Every level works this way.
The other structural fact is the reading section, which is a speed test as much as a comprehension test. Candidates routinely know the material and run out of time, and the failure looks like a vocabulary problem in the report and is not one. Practising reading for accuracy without a clock therefore prepares the wrong thing, and one timed section will tell you more than a month of untimed study.
The gap between the bands is also uneven and the largest jump is in the middle, where the vocabulary and the reading length both step up sharply. Planning linearly from the band below is the commonest way to arrive underprepared with a booking already paid for.
Getting extremely good at recognising Japanese
Here is the stall and the exam causes it without meaning to. Every hour spent on the syllabus is an hour on reading and listening, both of which are receptive, and both of which are the cheaper half of any language. Output never appears in the plan because it never appears in the test, and nothing in the system will report its absence.
The correction has to be a deliberate purchase rather than a study decision, and it has to be made early, because the gap widens with every year of receptive work. An hour a week with a person, from the first year, is the whole of it, and what that hour has to contain if it is not to become a pleasant chat is set out in finding someone to speak to, and what to do in the hour. The certificate is still worth having; it simply is not evidence of the thing you will be judged on in a room.
Two different measurements
What the paper measures against what a room measures
The pans are level because both are real. They are not level in what a certificate can tell anybody, which is the point.
- What the JLPT scoresReceptive, on paper, in a hall
- Vocabulary and grammar recognised
- Reading, under time pressure
- Listening, scripted and clear
- A section floor and an overall total
- What a workplace assessesProductive, in a room, in the first week
- Taking a turn in a meeting
- Choosing the right politeness level
- Writing an email somebody answers
- Following two people who interrupt each other
The band is evidence of the left pan and silent about the right
Nothing in the right hand pan is tested at any level, which is why it has to be bought separately and early. The exam is not wrong to omit it; a learner is wrong to let the exam decide what the year contains.
Two things to prepare with
The official sample papers are published free by the examining bodies, with the audio, and they are the only material guaranteed to match the current format at each level. They replaced the commercial preparation book, which reconstructs the paper and is usually one revision behind, and the published set is large enough to run a monthly mock for a year without repeating a paper.
The second is one reading section, sat under the clock, before any vocabulary study begins. It replaced doing the questions at your own pace, which measures none of the things this section measures. Free, it costs an hour, and it produces the only number that should decide how the remaining months are spent: whether you finished.
The mistake: choosing the level by your kanji count
Kanji is the visible measure, it is the one learners track, and it is the one that runs ahead, because characters can be drilled and listening cannot. A candidate who picks a level by how many characters they know has picked by their most measurable skill, and the listening section is the one that most often falls below its floor.
Sit a full paper, look at the section that is nearest the minimum, and spend disproportionately there. It will usually be listening, it will usually stay uncomfortable, and the alternative is a paper that reports a good overall score and a fail. That listening work is a different subject from the exam, and it is pitch accent, and speech that drops what is obvious.
What the bands stand in for
Everything the paper measures rests on characters learned inside words rather than as shapes, which is readings, radicals, and an order that makes them stick, and on a vocabulary split across three registers, in native words, Chinese readings, and English borrowings. The grammar section is largely about particles and endings, in particles, and a verb that waits at the end. And the thing the exam does not test at all is the one a workplace assesses in the first week, which is plain, polite and humble, and choosing between them.