Ten JLPT Exams, Two Certificates

In the post about my transition period after finishing my Bachelor’s, I mentioned passing the JLPT N3 in December 2018 and said I’d tell the language story properly in a separate post. That was a while ago. This is that post.

It took me this long partly because I kept assuming the story needed an ending, and the ending I had in mind was a certificate I still don’t have.

So here is the whole thing without one.

The Scoreboard, First

I’d rather put the numbers at the top than build up to them, because the numbers are the reason this post is worth reading. Most “how I passed the JLPT” articles are written by people who passed. Here is every result slip I have, section by section:

(At N4 the exam combines vocabulary, grammar and reading into a single 120-point section, which is why that row looks different.)

Read down the Reading column. I did that properly for the first time while writing this post, and it reorganised everything I thought I knew about my own Japanese.

Ten exams. Two certificates. A PhD earned in Japan somewhere in the middle of that column, and a full-time engineering job in Tokyo at the bottom of it.

Starting From Zero, and Signing Up Far Too Early

I started learning Japanese from zero while I was still an undergraduate at ITB, in Indonesia — no Japanese school, no year abroad, no family connection to the language.

Two or three months in, I registered for the N4, and sat it in December 2016 — around four or five months after I’d started.

That is, objectively, too early. I want to be fair to my past self about why I did it anyway: a registered exam date is the most effective study tool I have ever found. It converts a vague intention into a deadline with a fee attached. I don’t regret registering early. I just want to be honest that early registration buys you motivation, not readiness.

82 out of 180. Failed. And the interesting part wasn’t the fail — it was where it came from.

The slip grades your vocabulary, grammar and reading separately as a letter, and mine came back A, A and B. On the written half of that exam I was, by the JLPT’s own reckoning, in good shape. Then there was the listening section: 23 out of 60.

Everything I had been doing was silent. Textbooks, kanji drills, grammar exercises, all of it read off a page in a quiet room. Then the audio started and I discovered that I had been studying a written language that happens to also be spoken. My vocabulary was rated A and I still failed the exam, because one section of it was made of sound.

The Fix Was Idol Variety Shows

A friend gave me the advice that changed the whole trajectory, and it did not sound like study advice: watch Japanese idol variety shows. Specifically Nogizaka46 and Keyakizaka46.

I want to defend this properly, because it sounds like an excuse to watch television.

Language-learning audio is designed to be understood. One speaker at a time, clean studio recording, no crosstalk, vocabulary chosen for your level. Real Japanese is none of those things — and a variety show is the least of those things. Six people talking over each other. Jokes that depend on a word’s second meaning. Sentences abandoned halfway through because someone interrupted. On-screen text that comments on what was just said rather than transcribing it. Speed that nobody slows down for you.

It was, for a long time, incomprehensible. That turned out to be the point. Listening tests are not a test of whether you know the words; they’re a test of whether you can keep up while the audio moves on without you. Variety shows train exactly that, because they refuse to wait.

They also fixed the actual reason I wasn’t studying my listening: it was boring. This wasn’t. I would have kept doing this whether or not there was an exam, which made it the only part of my routine that never needed discipline.

If you take one practical thing from this post, take this one: your listening will improve at the speed of the material you enjoy, not the material you approve of.

Not Retaking N4, and Failing N3 Twice

I didn’t retake N4. Not because I thought I was above it — the reasoning was more boring than that. I was fairly confident that if I sat it a second time I would pass, and a result I already knew in advance wasn’t worth the fee or the months of waiting for the next sitting. The fail had already done its real job: it told me my listening was hollow, and by the time the next exam came around I’d spent that whole gap fixing precisely that. So I registered for N3 instead.

I sat N3 in July 2017 and failed that too — 82 out of 180, the same total I’d scored at N4. But the shape of it had completely inverted:

• Listening: 36 / 60 — my best section, at a level harder than the one where I’d scored 23.
• Vocabulary and grammar: 24 / 60. Reading: 22 / 60. Both now my worst.

Six months of variety shows had taken listening from the thing that sank me to the thing holding me up, across a level jump. Meanwhile the written half — the half I’d been rated A on at N4 — had fallen behind, because N3 asks considerably more of it and I’d spent those six months watching television.

That flip is the single most useful thing the JLPT ever did for me, and it’s the argument for sitting exams before you’re ready. A failed exam is not just a rejection; it’s a diagnostic report on where your Japanese actually is, as opposed to where you feel it is. My variety-show habit had quietly rebuilt one skill while I neglected the two that a textbook is genuinely good at teaching. Without the fail I’d have kept optimising the thing I had already fixed.

Then came the third exam and the one I still think about.

December 2017. 94 out of 180. The pass mark for N3 is 95.

I failed by one point.

There is no lesson in that paragraph. I’ve tried to find one several times. You can’t appeal it, you can’t argue that you were close, and the exam has no interest in how many times you have already tried. The score is the score. What I remember is how strange it felt to be genuinely sad about a number while also knowing there was nothing whatsoever to say about it.

You just register again.

Passing N3, One Year Later

The next sitting after that one-point fail was July 2018, and as best I remember I simply didn’t turn up for it. That was the summer of my undergraduate final project — Tugas Akhir 1 and 2, and the defense that ends a physics degree at ITB. There was no room in it for a language exam, so the language exam didn’t happen.

I mention it because it’s the first time this decision shows up in my record, and it is not the last. Twice now a thesis defense has landed on top of a JLPT sitting, and both times the defense won. That is the correct answer both times.

So: December 2017, then December 2018. I passed N3 at the fourth JLPT sitting of my life — 115 out of 180, and this time listening was 45 out of 60. The section that had scored 23 two years earlier finished as my strongest by a clear margin.

It landed in a busy month. I was finishing my run as Coordinator of Teaching Assistants for the Advanced Physics Lab, my Bachelor’s already behind me, and the path to Japan not yet certain. Just over a year later I flew to Tsukuba.

The N2 I Failed Before I Ever Left Indonesia

Seven months after passing N3, still in Indonesia, I sat N2 in July 2019.

66 out of 180. Vocabulary and grammar 20, reading 18, listening 28. The pass mark is 90. It is comfortably the worst result on my record — worse, in raw terms, than any of my N1 attempts, and I had been studying Japanese for three years by then.

Note the reading: 18. Every section of the JLPT also carries a minimum you have to clear on its own, and at this level that minimum is 19. So even if my total had somehow reached 90, that 18 would have failed me anyway. I didn’t know that at the time, and it’s the first appearance of a problem I then ignored for another six years.

I’d like to tell you there’s a dramatic reason. There isn’t one. That was the summer everything else was happening: applications, English tests, an entrance exam in Japan in August, an entire departure to organise. N2 was one more thing in a stack of things, and it got what was left over, which was nothing.

Remember this score. It becomes the most useful number in this whole post four years later.

Four Years in Japan, and No Exam at All

Then I arrived in the country the language belongs to, and stopped taking its exam for four years.

This is the part I’d do differently. Between the Master’s and the PhD there was always a research reason to postpone the JLPT — a seminar, a paper, an experiment that couldn’t move. The lab will always give you one. I’ve written about what that cost me when job hunting started and I needed a language line on my rirekisho: the certificate you don’t have is the one that can’t get you past document screening, and the hiring calendar does not wait for your results to arrive.

Take the JLPT earlier than your research schedule wants you to. I say that as someone who didn’t.

The Same Exam, Four Years Apart

July 2023, second year of the PhD: I passed N2 with 102 out of 180.

Not my first N2. My second — and that is what makes it worth writing down.

I have to be honest about how it happened, because it’s the least earned certificate I hold: I don’t really remember studying for it. There was no routine, no plan, no drill schedule. There wasn’t time for one. It felt like N3 plus three years of simply living in Japan — the lab, the ward office, the supermarket, the landlord, the bank counter, the entire ordinary business of existing somewhere in a language.

And that is exactly why the 2019 attempt matters. Put the two side by side:

Same exam. Same person. Roughly the same amount of deliberate preparation, which is to say almost none either time. Thirty-six points apart, and the only variable that really changed was that I had been living in Japan for three years.

But look at where the 36 points came from. Listening went up by 21 — it did nearly two-thirds of the work on its own. Vocabulary and grammar gained 11. Reading moved four points, from 18 to 22.

I passed N2 on the back of a listening score of 49 out of 60. My reading was three points above the sectional minimum. At the time I read that result as “my Japanese got better,” which was true but far too generous to myself. What actually happened is that the part of my Japanese which immersion feeds got dramatically better, and dragged an otherwise mediocre exam over the line.

That is as close to a controlled experiment on immersion as I’m ever going to run on myself, and it comes out unambiguously: living in Japan will upgrade your Japanese for you. Immersion is not a myth. You will get better without deciding to.

It’s also, as I was about to find out, a promise with a ceiling.

The N1 Wall

Three attempts. Three fails. 74, then 53, then 73.

The pass mark for N1 is 100 out of 180 overall, plus a minimum in each section — so 74 is not a near miss. It’s a clear, unambiguous “not yet,” delivered three times.

Let me take them individually, because they are not the same failure.

July 2024 — 74/180. Vocabulary and grammar 24, reading 19, listening 31. Final year of the PhD, job already secured. This is my honest baseline: what N1 looks like when I sit it as the person daily life in Japan has made me, with some preparation but no campaign. Twenty-six points short — and reading landing exactly on the sectional minimum, not one point of margin.

December 2024 — 53/180. Vocabulary and grammar 12, reading 8, listening 33. The lowest score I have ever received on any exam, and the one I’m least bothered by, because it wasn’t a collapse. It was a decision.

That was the month of my first doctoral defense. I had registered for N1 long before I knew how that month would look, and when it arrived there was no version of it where I prepared for both. So I didn’t. I put everything into the defense and I sat the JLPT anyway, with no preparation at all, because the seat was paid for and there was no reason not to.

53 out of 180 is exactly what that costs. I knew roughly what it would cost when I chose it, and I would choose it again — a doctoral defense is not repeatable in the way a language exam is. The JLPT runs twice a year, every year, forever. The defense was in December.

Six years earlier, facing the same collision with my undergraduate defense, I made the same call and skipped the sitting entirely. The only thing that changed in between is that this time I sat the exam anyway and let the score record the decision.

I’m spelling this out because a score chart like mine invites a dramatic reading — the burnt-out student, the toll of overwork — and that reading would be false. Nothing broke. I chose which one mattered that month, and I chose right.

And that unprepared exam turned out to be the most informative one I have ever sat, for a reason I only saw years later. Put it next to July, five months earlier:

• Vocabulary and grammar: 24 → 12. Halved.
• Reading: 19 → 8. Less than half.
• Listening: 31 → 33. It went up.

Take away all preparation, and the studied skills fall off a cliff while the skill built by living here doesn’t even notice. Listening isn’t something I revise; it’s something I own. On that score sheet, everything else turned out to be rented.

December 2025 — 73/180. Vocabulary and grammar 20, reading 13, listening 40. This is the one that actually stings, and it’s the reason this post exists.

By then I had graduated, moved to Tokyo, and started full-time work as an engineer. I speak Japanese all day. The training was in Japanese, the daily reports were in Japanese, the materials were in Japanese. On paper this should have been my strongest attempt yet.

It was one point below my first — while my listening reached 40, my best N1 listening by a distance. Something had improved substantially in those eighteen months. It just wasn’t the thing the exam weighs most heavily.

Why Living in Japanese Doesn’t Get You N1

That 73 forced me to admit something I had been comfortably avoiding: N1 is not a test of whether you can live in Japanese.

I can. I do it every day. I can hold a technical discussion, write a daily report, argue with an administrative process and win. None of that is what N1 measures.

The Japanese I use is my Japanese — the vocabulary of my field, my routine, my colleagues, my paperwork. It’s deep in a narrow channel. N1 is the opposite shape: deliberately broad, formal, literary, and full of constructions that a working adult can go years without meeting once. Newspaper editorials. Grammar that exists to be recognised rather than spoken. Reading passages long enough that the real test is stamina.

Living here builds the channel deeper. It does not make it wider. And the exam only rewards width.

So the pattern I met at N3 came back one level up, with the same shape and a bigger gap: at some point immersion stops carrying you, and the only thing that works is the unglamorous, deliberate, sit-down-and-study kind of study — which is precisely the thing that a full-time job and a doctorate are best at eating.

That was my theory, anyway. Then I pulled every score sheet I own into one table for this post — something I had never done in ten years of sitting this exam — and the theory turned out to be right, but vague. The numbers name the section.

It’s reading.

Here is my reading score at every level since N3, in order:

22 · 32 · 37 · 18 · 22 · 19 · 8 · 13

Everything left of the gap is N3. Everything right of it is N2 and N1, and it has not been above 22 out of 60 in seven years. My listening over the same stretch went 36, 34, 45, 28, 49, 31, 33, 40. One of those columns is a story of steady improvement. The other has been flat on its back since 2019, and I had genuinely never lined them up before.

It gets worse, and more useful. Each section carries its own minimum — 19 out of 60 — that you must clear regardless of your total:

• July 2024: reading exactly 19. On the line, no margin.
• December 2024: reading 8. Below the minimum.
• December 2025: reading 13. Below the minimum.

So for two of my three N1 attempts, the total score was never the real story. Even if I had somehow found the missing 27 points elsewhere, that reading section would have failed me on its own. I have spent three attempts thinking of myself as someone who is short on points. I am actually someone who is short in one specific section, and no amount of general improvement will fix that.

This is the part I find hardest to write. The habit that carried me from N4 to N3 was reading the diagnosis and aiming at the weak section — listening at 23 sent me to the variety shows; reading at 22 sent me back to the textbook. Somewhere after that I stopped doing it. I kept sitting the exam, kept noting the total, kept concluding “not enough yet,” and never once opened the breakdown that was sitting there telling me exactly where to aim.

Ten years of results, and the single most useful thing in them was something I could have read in five minutes at any point since 2019.

Attempt Four

I sat N1 again on the morning of 5 July 2026, in Tokyo. Attempt number four.

It felt better than December. I want to be careful with that sentence, because “it felt better” is the least reliable data in existence — I have felt fine walking out of exams I failed, and I sat the 53 knowing exactly what was coming. The result arrives around late August. It may well be another one.

I registered anyway, and I’ll register again after that if I need to. Not out of stubbornness exactly. It’s more that I’ve been here before: four exams to clear N3, and the fourth one worked.

What’s different now is that I sat that exam in July without knowing what I’ve just spent this post working out. If the result in August is another fail, it will be the first one I’ve received with a plan attached to it — and the plan is embarrassingly simple, which is usually a good sign. Read more. Read things I would never choose. Read them until the reading column stops being the one I hope nobody looks at.

For what it’s worth, I’m not even doing this one alone — I’ve been working through beginner Korean and Chinese alongside it, which is either a terrible idea for my N1 score or the reason any of this is still enjoyable after all these years. Possibly both.

What I’d Tell Someone Starting Now

1. Register before you’re ready. A booked exam date is worth more than a study plan. Just be clear-eyed that early registration buys motivation, not a pass.
2. Treat a fail as a diagnostic, not a verdict — and actually read the diagnosis. My N4 slip said listening 23/60; my first N3 slip said listening 36/60 and reading 22/60. Those two pieces of paper told me exactly what to do next, twice. Then I stopped reading them for seven years and wondered why I kept failing. Put every result you’ve ever had in one table. It takes five minutes and it will tell you something your totals never will.
3. Check the sectional minimums, not just the total. Twice I failed N1 with a reading score below the per-section floor, which means the points I was chasing elsewhere would not have saved me. A section below the minimum is a different problem from a total that’s too low, and it needs a different fix.
4. Fix your weakest skill with material you actually enjoy. Variety shows beat listening drills because I kept doing them. 23 to 36 across a level jump, in six months, from something I’d have watched anyway. The best method is the one that survives a bad week.
5. Don’t wait for a quiet semester. There isn’t one. I postponed the JLPT for four years in Japan waiting for research to calm down, and it never did.
6. Immersion buys you listening. It does not buy you reading. Between my two N2 attempts — Indonesia to Japan, no real study either time — listening gained 21 points and reading gained four. That’s the whole shape of immersion in two numbers. Living here will carry you a long way for free, and it will carry you almost entirely through your ears.
7. When two things collide, choose, and accept the price. I skipped a sitting for my undergraduate defense and scored 53 during my doctoral one. Both were the correct call — a defense happens once, the JLPT runs twice a year forever. What you can’t do is half-prepare for both and be surprised by the outcome of either.
8. Failing repeatedly is not a disqualification from anything. I hold a doctorate from a Japanese university and work as an engineer here in Japanese. My N1 certificate does not exist. Both of those are true at once, and the internet is unusually bad at telling you that.

Check the Current Rules

The JLPT’s levels, pass marks, sectional minimums, registration windows and fees change over time, and they differ between test sites in Japan and abroad. Every score above is copied from my own result slips, sat between December 2016 and July 2026 in Indonesia and Japan. The thresholds I planned around were 95/180 for N3, 90/180 for N2 and 100/180 for N1, each with a minimum of 19/60 in every section on top of the total — which is why a good total can still fail, and why two of my N1 attempts were beaten by one section rather than by the total. Confirm the current rules for your own level and country on the official JLPT site (jlpt.jp) before you register, and don’t build a study plan on one candidate’s numbers.