Every product that teaches Japanese counters competes on the same number: how many counters it contains. One app ships 84. A well known reference list runs to 350. The dedicated practice sites organise theirs into packs and levels, and the marketing copy always implies the same thing, which is that the counters you have not met yet are the problem.
They are not the problem. I am a native speaker, I live in Tokyo, and I want to tell you something that the app store listings will not: the number of counters a Japanese adult actively produces in an ordinary week is small enough to fit on the back of a receipt. Tofugu, which built the most complete counter reference in English, says so itself — depending on your goals you can get away with memorising two, nineteen, or sixty-six of them out of the 350 they catalogued. Their top tier, the one they label absolutely must-know, contains two counters.
So an app with 84 counters and a spaced repetition scheduler is not solving your problem. It is solving a problem you do not have, thoroughly, on a schedule, forever.
This article is about what the real problem is, which tool actually drills it, and how little time you should spend here. It is the eleventh article in a series where I look at Japanese learning products from the inside of the language, and it reaches a conclusion I have not reached in any of the previous ten: counters matter less than you think, and I am going to talk you out of most of the work.
What actually goes wrong, from the listening side
When a learner counts something in front of me, there are two separate things that can break, and they break in completely different ways.
The first is choosing the counter. You have three pencils. Pencils are long and thin, so the counter is 本. You have three sheets of paper, which are flat, so the counter is 枚. This is the part every app teaches, every blog post lists, and every learner worries about.
The second is the sound change. Having chosen 本, you now have to say it, and the number in front of it changes shape. One is いっぽん, three is さんぼん, six is ろっぽん, eight is はっぽん, ten is じゅっぽん. The counter did not stay put. It moved between h, b and p depending on which number it was standing next to, and the number itself grew a small っ in some cases and not others.
Here is the thing I would like you to sit with. The second problem is the one that actually stops learners mid-sentence, and it lives entirely inside the counters they already know. Nobody freezes because they failed to retrieve the counter for a suit of armour. They freeze because they know 本 perfectly well, they have the number six in their head, and the bridge between them collapses.
An app that adds counter number 85 to your review queue has done nothing about this. An app that drills the sound changes on the five counters you use every day has done everything.
If the small っ in いっぽん feels like a detail, it is not. It is a full beat. Japanese runs on mora timing rather than stress timing, and that っ occupies one mora of silence exactly as a spoken syllable would. Saying いっぽん with the っ collapsed is the same category of error as saying おばさん when you meant おばあさん. I wrote about that clock in detail in my guide to why you can understand Japanese but cannot speak it, and counters are where it bites soonest.
The direction problem, which is why lists do not work
There is a structural reason that counter lists fail learners, and once you see it you cannot unsee it.
A counter list is indexed the wrong way round. It is organised counter first: 本 is for long thin things, 枚 is for flat things, 匹 is for small animals. That is a perfectly good structure for reading. You meet 三匹 in a sentence, you look up 匹, you learn it means small animals, you move on.
But production runs in the opposite direction. In real speech you do not start with a counter. You start with an object — a rolled up poster, a slice of ham, a pair of headphones, a tube of glue — and you have to arrive at a counter. You cannot look up an arbitrary object in a list that is indexed by counter. The lookup you need is object to counter, and the resource you have been given is counter to objects.
This is the same shape of failure I keep finding across this whole series. Products hand you the material and never make you perform the retrieval in the direction that speech requires. I found it with pitch accent, where two products display the pitch pattern but no product asks you to produce it. I found it with politeness and register, where every product teaches both forms and none asks you to choose between them for a given listener. Counters are the cleanest example yet, because the direction of the index is visible on the page.
How I judged these
Three tests, in order of how much they matter.
1. Does it ever go from object to counter? Does the exercise show you a thing and require you to supply the counter, or does it show you a counter and require you to supply information about it? Only the first is production.
2. Does it drill the sound changes, or merely list them? Almost every resource prints the いっぽん さんぼん ろっぽん table somewhere. Printing it is not drilling it. The test is whether the product forces you to generate an irregular combination under time pressure, repeatedly, until it is automatic.
3. Does it tell you when to stop? A counter tool that does not tell you that nineteen counters is a reasonable finish line is selling you an infinite queue. This one matters more than it sounds, because the opportunity cost is real: the hours you spend on counter 60 through 84 are hours not spent on things that actually damage your Japanese.
The comparison
| Tool | Cost | Object to counter? | Drills sound changes? | Tells you when to stop? | Can it pay me? |
|---|---|---|---|---|---|
| 助数詞を練習 / jyosuushi.com | Free | Yes. The prompt is an English phrase and you type the answer | Yes, because you must type the full combination | Partly. Packs are bounded by level | No |
| Tofugu counter list | Free | No. It is a reference | No | Yes, explicitly, in tiers | No |
| Kaunta | $5.99 one time | Unclear from the listing | Yes. It has a dedicated Exceptions Mode | No. It ships 84 | No |
| WaniKani | $9 per month and up | No. Counters arrive as vocabulary | Partially, as separate vocabulary items | No | No |
| Bunpro | About $5 per month | No. Counters are not in the grammar library | No | n/a | No |
| JapanesePod101 | Subscription | No. The lessons explain | No | No | Yes, 25 percent |
| Rocket Japanese | About $150 and up | No. I could not find a counters lesson at all | No | n/a | Yes, 40 percent — I am not in it |
| Anki | Free | Only if you build it yourself | Only if you build it yourself | You decide | No |
Read the last column against the third. The only tool here that drills counters in the direction speech requires is a free website made by a hobbyist, and it pays me nothing. The two products with the biggest commissions here are the two that barely touch counters at all. I did not arrange that. It is simply what I found.
What I actually checked, and what I did not
I read the published lesson lists, store listings and grammar libraries directly rather than trusting review sites. Where a product sits behind a paywall I have said so instead of guessing.
I have not installed Kaunta, because it is iPhone only and I did not want to describe an exercise loop I had not seen. Everything I say about it comes from the App Store listing, and where the listing is ambiguous I have marked it ambiguous rather than filling the gap. That matters here specifically, because whether Kaunta asks object to counter or number to reading is the single most important fact about it, and the listing does not settle it.
I also could not load practice-japanese.com, which returned an error to me, so it is absent from the table rather than judged.
The tools, one by one
助数詞を練習, at jyosuushi.com — the best thing here, and it is free
What it does: a quiz site built around counters. You pick a study pack, and the prompt appears in English — something like 19 coworkers or 5 straws. You then type the whole number and counter combination in hiragana. It gives you feedback and information about that combination afterwards. There is an Explore tab for studying before you drill, packs for N5 counters, N4 counters, Essential and Common, and controls for skipping a question or marking an honest typo so it does not poison your record.
Why it wins: look at that prompt format again. 19 coworkers. You are given the object and the quantity, and you must produce the counter and survive the sound change, by typing, with no options to pick from. That is the exact retrieval that speech demands, and after eleven articles of looking for it in this market, this free site is where I found it done properly.
The English prompt is doing quiet work too. Because the object arrives in English, the site cannot leak the answer to you the way a Japanese prompt would. You have to categorise the object yourself, which is the skill.
Cons: it is in beta and says so. The generator can produce prompts that no human would ever say, and the batches are not customisable, so you cannot yet drill exactly the six counters you care about. The interface is plain.
Who should use it: everyone reading this article. Start here, and in most cases stop here.
The Tofugu counter list — the reference, and the permission to stop
What it does: 350 counters, sorted not alphabetically but by how useful they are. The tiers are Absolutely Must-Know at 2, Must-Know at 17, Common at 47, Somewhat Common at 205, Rare But Interesting at 22, and Gairaigo at 57. It is free, and there are spreadsheet versions floating around.
Why it matters: the sorting is the product. Every other counter resource presents a flat list and leaves you to infer priority, which you cannot do, because you are the one person in the room who does not yet know which counters are common. Tofugu did the triage and published it.
And it is the one resource that gives you explicit permission to stop. Two, nineteen or sixty-six, depending on what you want. When you next feel anxious about a counter you have never seen, go and check which tier it is in. If it is in the 205, you have found a counter that I would also have to think about.
Note that Tofugu is a competitor of mine, not a partner. I am pointing you at their reference because it is the best one and there is no version of this article that is honest and leaves it out.
Kaunta — the honest paid pick, if the sound changes are your wall
What it does: an iPhone app from PRPL development AB. Free to start with 21 counters and unlimited drilling on them, then a one time upgrade at $5.99 for lifetime access to all 84. Reviews are scheduled with FSRS spaced repetition. You answer by typing in Japanese or by choosing from multiple choice, and it accepts hiragana, katakana and romaji. There are three modes: Smart Mode for weak spots, Random Mode, and Exceptions Mode, which targets the irregular readings.
Why I am recommending it despite my own argument: Exceptions Mode is aimed squarely at the thing I said matters most. A dedicated drill for irregular number and counter combinations is exactly right, and no other product in this comparison has one. If you know your counters and still stall on ろっぽん, this is the tool built for you, and $5.99 once with no subscription is an honest price.
Where I would push back: the 84 counters are the headline, and the 84 counters are the least valuable thing in the box. The free tier at 21 counters is, by Tofugu triage, already past the Must-Know tier of 17. Many people should use the free tier permanently and never pay, and I would rather say that plainly than pretend the upgrade is necessary.
Unverified: I could not determine from the listing whether the exercises run object to counter or number to reading. If it is the latter, it fails my first test entirely and is a sound change tool only. Worth thirty seconds on the free tier to check before you decide.
WaniKani — counters arrive, but as vocabulary
WaniKani teaches counters, and it teaches them well within its own logic. The counter 〜台 appears at Level 3 as a vocabulary item meaning big machine counter, and you are tested on its meaning and its reading, with a note reminding you to omit the placeholder from your answer.
But that is a vocabulary card. You are never shown a car and asked which counter it takes. The information arrives, the retrieval direction is recognition, and the counters are scattered across levels according to the kanji schedule rather than gathered by usefulness. WaniKani remains the single best product I recommend in this whole series, as I explain in my main comparison of Japanese learning apps, and it is not trying to be a counter tool. Do not buy it for this.
Bunpro — it simply does not cover them
This surprised me. Bunpro is the product I trust most on grammar, and counters are not in its grammar library. Learners have been asking for them in the community forum as a feature request. What exists is the pattern 何 plus counter plus か, which is a question construction rather than a treatment of the counter system.
I am noting this because Bunpro tops several of my other articles, and it would be easy for a reader to assume it covers everything. It does not cover this.
JapanesePod101 — good explanations, no production
There is a real counter series here: an Introduction to Japanese Counters, a lesson devoted to 個, Basic Japanese lessons 14 and 15 on general counters, and an appendix on counters. For explanation this is solid, and the audio means you actually hear the sound changes spoken rather than reading them in a table, which is genuinely valuable.
But it is audio and video. Nothing in it requires you to produce a counter for an object you were not prepared for. You listen, the hosts explain, you agree. This is the same limit I described when I looked at how textbook Japanese differs from the real thing, and it applies here unchanged. JapanesePod101 pays me 25 percent and it is still not the tool for this job.
Rocket Japanese — numbers yes, counters no
Rocket has a lesson called Counting in Japanese covering one to ten, and a lesson called Japanese Numbers covering one to one hundred plus age and money. Searching the published course outline, I could not find a lesson on the counter system at all.
I want to be careful here rather than sweeping. I have not seen the inside of the paid course, so the honest statement is that counters do not appear in the published outline, not that they are definitively absent. But Rocket pays 40 percent, the highest rate of any product on this page, and on this topic I cannot point you at it.
Anki — the do it yourself option
Free, and you can build exactly the deck this article argues for: object on the front in English, number plus counter in hiragana on the back, restricted to your twenty counters, weighted towards the irregular combinations. Nobody sells that deck because there is no money in a deck of twenty items. If you already run Anki, an evening of deck building will beat every paid option here.
What a native speaker actually does, which nobody teaches
Now the part that no product will tell you, because no product can sell it.
We have an escape hatch, and we use it constantly. The traditional numeral series — ひとつ, ふたつ, みっつ and so on up to とお at ten — attaches to almost anything. When I am holding an object whose counter I do not immediately know, or whose counter feels stiff for the situation, I use つ. Beyond ten, 個 does the same job. Both are completely ordinary adult Japanese.
This changes the strategy entirely. You do not need coverage across 84 counters, because you need about twenty counters said confidently plus a fallback that is always available. The fallback is not a beginner crutch that you are supposed to grow out of. It is what fluent speakers do.
Two constraints worth knowing. The つ series stops at とお, so it will not carry you past ten. And 人 for people is irregular in exactly the places you need most, since one person is ひとり and two people is ふたり before the pattern becomes regular at さんにん. Those two irregular forms are worth more drilling than the next fifty counters combined.
There are also counters that I would get wrong. Traditional Japanese assigns specialised counters to all sorts of things — a wardrobe chest, a butterfly, a mirror — and I would not reliably produce them. Rabbits are the famous case, counted with 羽, the counter for birds, apparently because categorising them as birds sidestepped an Edo period prohibition on eating four legged animals. It is a wonderful piece of trivia. It is also not something you need, and when I want to count rabbits I say 匹 or つ like everybody else.
If an app is drilling you on counters that native speakers avoid, it is not making you sound more Japanese. It is making you sound like a person who has been studying a list.
And in casual speech, 個 is eating the system. In a convenience store you will hear 個 attached to things a textbook would assign 本 or 枚. The formal system is intact in writing and in careful speech, and it is quietly simplifying in the mouths of everyone under fifty.
Why no product drills this properly
The same reason I found behind every other gap in this series, and it is worth stating plainly.
Counters appear on the JLPT. For N5 you want a core set of roughly eight to ten, and 人 and 本 show up most often. But every question on the JLPT is four option multiple choice, so the exam asks you to recognise a counter, never to produce one. There is no speaking section and no writing section at any level, which I go through in detail in my article on what the JLPT actually measures.
So the only external standard in this market rewards recognition and coverage. Coverage is countable, which makes it marketable — 84 counters is a feature you can print on a store listing. Object to counter production under time pressure is neither countable nor marketable, and the one place I found it done properly is a free beta site with no business model.
The honest ranking: counters are lower priority than you think
Here is where this article breaks from the other ten, and I want to be direct about it.
In every previous article I found a gap and told you it mattered more than the products admit. This time the gap is real and it matters less.
The reason is the cost of the error. If you use the wrong counter, the sentence is understood instantly and completely. You say 三本 for three apples, I hear three apples, we continue. It reads as slightly foreign in a way that is entirely unremarkable in Tokyo, and — this is the important part — it is a visible error, so you find out and fix it.
Compare that with getting register wrong. Casual speech aimed at the wrong person is not misunderstood either, but nobody corrects it, and the cost is charged silently against how you are perceived. That asymmetry is the whole subject of my article on casual and polite Japanese, and it is why register belongs near the top of your priorities and counters do not.
An experienced learner in the Bunpro forum put it more bluntly than I would have dared: counters are one of those things you should not bother actually learning, in the sense of formal study, because they come with exposure. I think that is slightly too strong — the sound changes genuinely do need drilling — but the instinct is correct, and it is the opposite of what an app with a review queue needs you to believe.
What to actually do
The thirty minute version, which is most of the value available here.
- Open the Tofugu list and read the top two tiers only. Two counters, then seventeen. That is your whole vocabulary target. Ignore the other 331 for now.
- Spend twenty minutes on jyosuushi.com, on the Essential pack, typing answers rather than skipping. Do this two or three times in the first week and then roughly monthly.
- Drill the irregular combinations specifically — いっぽん さんぼん ろっぽん はっぽん じゅっぽん and the same pattern on 匹 and 杯, plus ひとり and ふたり. Out loud, not silently. If this is the part that breaks, the free tier of Kaunta with its Exceptions Mode is worth the download, and possibly worth $5.99.
- Adopt the fallback deliberately. Decide now that you will say つ or 個 when you do not know, and stop treating that as failure. It is what I do.
- Stop. Do not build a counter habit. Counters are a thing you handle in a month and then acquire passively by reading and listening for the rest of your life.
The bottom line
The best counter tool in this market is a free beta website that asks you to type 19 coworkers in Japanese. The best reference is a free list published by a company that competes with me. The best paid option costs $5.99 once, and most readers should use its free tier and never upgrade.
The products with the biggest commissions — JapanesePod101 at 25 percent, which does pay me, and Rocket at 40 percent, whose programme I am not in — either explain counters without drilling them or do not cover them in their published outline at all. I have no financial reason to write this article and I am writing it anyway, because the alternative is that you spend three months in a review queue of 84 counters while your register, your pitch and your listening go unattended.
Learn nineteen counters. Drill the sound changes until they are automatic. Say つ when you are stuck. Then go and do something that actually changes how you sound.