Japanese Verb Conjugation: The Three Groups Explained

🗣️ Conjugation is a production skill, not a reading skill — you learn it by saying the forms, not by looking at them. SoraSpeak gives you unscripted spoken practice with AI Japanese teachers, plus instant corrections on grammar, particles and politeness explained in simple English. Free daily tier, in your browser. Start a free speaking session →

Here's the short version: Japanese has three verb groups, and once you know which group a verb belongs to, every one of its forms is predictable. Group 1 (godan) changes the final vowel. Group 2 (ichidan) just drops る. Group 3 is two irregular verbs, する and 来る, that you memorise once and reuse forever. There is no gender, no person, no plural, and no irregular past tense list of 200 items. Japanese conjugation is genuinely simpler than English — the difficulty is producing it at conversation speed, which is a different problem from understanding it.

This guide is built for that second problem. Tables first, then the order to learn the forms in, then drills you say out loud.

In this guide: the three groups · telling them apart · master table · what each form does · learning order · drills · suru compounds · the bottom line

The three verb groups in thirty seconds

Every Japanese verb in the dictionary ends in the vowel sound u: く ku, す su, つ tsu, ぬ nu, ぶ bu, む mu, る ru, ぐ gu, う u. That is the only thing they all share. What splits them is what happens to that final sound when you conjugate.

Group 1 — godan (五段), the u-verbs. The final u sound shifts across the vowel row: 書く kaku becomes 書きます kakimasu, 書かない kakanai, 書ける kakeru, 書こう kakō. Five different vowels, which is exactly what 五段 ("five levels") means. This is the biggest group by a wide margin — roughly three quarters of common verbs.

Group 2 — ichidan (一段), the ru-verbs. Drop る, attach the ending. 食べる taberu becomes 食べます tabemasu, 食べない tabenai, 食べて tabete. One level, no vowel shifting, no exceptions inside the group. If you can identify a Group 2 verb, you already know all of its forms.

Group 3 — irregular. Two verbs: する suru (to do) and 来る kuru (to come). That is the entire list of irregular verbs in Japanese. Compare that with English, which has over 200. する is also the most productive verb in the language, for reasons covered further down.

Textbooks disagree on the names. Genki says Group 1 / Group 2 / Group 3, Minna no Nihongo says I / II / III, linguists say godan / ichidan / irregular, and English-language resources often say u-verb / ru-verb. They are the same three categories. Pick one vocabulary and stop worrying about it.

How to tell godan and ichidan apart

This is the only genuinely tricky part of Japanese conjugation, and it takes about ten minutes to master.

Rule 1: if it does not end in る, it is Group 1. 書く, 話す, 飲む, 待つ, 買う, 泳ぐ, 遊ぶ, 死ぬ — all Group 1, no checking needed. That single rule sorts more than half of all verbs instantly.

Rule 2: if it ends in る preceded by an a, u or o sound, it is Group 1. ある aru, 分かる wakaru, 作る tsukuru, 乗る noru, 降る furu, 取る toru, 終わる owaru. Also no exceptions.

Rule 3: verbs ending in -iru or -eru are usually Group 2 — but not always. 食べる taberu, 見る miru, 起きる okiru, 寝る neru, 教える oshieru, 開ける akeru, 出る deru. These are the real Group 2 verbs, and they are the majority of the -iru/-eru set. The problem is the minority.

The ru verb traps

A small closed list of common verbs ends in -iru or -eru and conjugates as Group 1 anyway. These produce more beginner conjugation errors than everything else combined, because they look like the easy group and behave like the hard one.

The cruelty of the list is that three of them have Group 2 homophones, so the same sound conjugates two different ways depending on the kanji:

• 切る kiru (cut, Group 1) → 切ります · 着る kiru (wear, Group 2) → 着ます
• 帰る kaeru (go home, Group 1) → 帰ります · 変える kaeru (change, Group 2) → 変えます
• 要る iru (need, Group 1) → 要ります · 居る iru (exist, Group 2) → います

The test that always works

Forget the rules and use the polite form as a diagnostic. If ます attaches after an i sound that was not in the dictionary form, it is Group 1. If ます attaches where る used to be, it is Group 2.

• 帰る → 帰ります — an り appeared → Group 1
• 食べる → 食べます — る simply vanished → Group 2

The practical consequence: learn every new verb as a pair, dictionary form plus ます form. 帰る・帰ります. 食べる・食べます. It costs one extra syllable at the moment you meet the word and removes the guessing forever. Any decent dictionary marks the group; if you learn from listening instead, you will hear the ます form constantly and absorb the split without trying.

The master conjugation table

Here is one Group 1 verb, one Group 2 verb and both irregulars, across every form a speaker needs. Everything else in Japanese conjugation is built on top of these.

Note the reading trap in the last column: 来る is kuru, 来ます is kimasu, 来ない is konai, 来い is koi. Same kanji, three different readings across four forms. It is the single most irregular thing in the language and it affects exactly one verb.

The て and た sound changes for Group 1

Group 2 and Group 3 need no help here — 食べて, 食べた, して, した, 来て, 来た. Group 1 has a five-row sound-change table that is worth memorising as a block, because て and た share it exactly:

One exception exists in the whole table: 行く iku (go) → 行って itte, 行った itta, not 行いて. Learn it now and you are done with て-form exceptions permanently.

Potential and the ra-nuki question

Group 1 potential is a clean vowel change: final u → eru. 書く→書ける kakeru, 飲む→飲める nomeru, 話す→話せる SoraSpeakru, 帰る→帰れる kaereru. Group 2 replaces る with られる: 食べる→食べられる taberareru, 見る→見られる mirareru.

In real speech, many Japanese speakers drop the ら from Group 2 potential forms: 見れる mireru, 食べれる tabereru, 来れる koreru. This is called ら抜き言葉 (ra-nuki kotoba, "ra-dropped speech"). It is genuinely common — you will hear it constantly on television and from anyone under forty — and it is still non-standard in writing, marked wrong in exams, and avoided in business Japanese. The honest advice: produce the full form, recognise the short one.

There is also a practical reason the short form exists. 食べられる is simultaneously the potential ("can eat") and the passive ("be eaten"), so 食べれる removes an ambiguity. That is a real function, not laziness, which is why the form keeps spreading despite disapproval.

One more note: potential verbs take が, not を. 日本語が話せます (nihongo ga SoraSpeakmasu) — "I can speak Japanese." If that surprises you, our Japanese particles guide explains why ability words attract が.

A warning about the imperative

書け, 食べろ, しろ, 来い are grammatically correct and socially dangerous. The plain imperative is rough — it belongs to sports coaching, road signs, military orders, angry parents, and men shouting at each other in fiction. Used at a colleague or a stranger it is close to a threat.

What you actually want in almost every situation is the て form plus ください: 書いてください (kaite kudasai), 食べてください (tabete kudasai), 来てください (kite kudasai). Learn the imperative to understand it when you meet it, not to use it.

What each form actually does when you speak

A conjugation table is a list of shapes. Here is what each shape buys you in a live conversation.

ます — politeness with zero risk. The default register for anyone you do not know well. It is never wrong in the way plain form can be wrong, so it is the correct place for a learner to live.

て — the connector, and the most valuable form in the language. On its own it links clauses: 起きて、朝ごはんを食べて、会社に行きます (okite, asagohan o tabete, kaisha ni ikimasu) — "I get up, eat breakfast and go to work." It is also the required base for a long list of everyday patterns: ている (ongoing action), てください (please do), てもいい (may I), てから (after doing), てみる (try doing), てしまう (do completely, or regrettably). Master て and roughly a third of beginner grammar comes free.

た — past, and much more. Past tense, obviously, but also 〜たことがある (have ever done), 〜たほうがいい (had better), 〜たら (if/when). It is the same sound change as て, so learning them together halves the work.

ない — negation, and the base for more negation. 〜ないでください (please don't), 〜なければならない (must), 〜なくてもいい (don't have to).

Potential — the single most useful form for a learner talking about themselves. 日本語が少し話せます. 明日は行けません. It is how you describe your own limits, which is most of what a language learner needs to say.

Volitional — invitations and intentions. 行きましょう (let's go) is the polite workhorse. The plain 行こう is what friends say to each other. Add と思います to the plain form and it becomes a stated plan: 来年日本に行こうと思います (rainen Nihon ni ikō to omoimasu) — "I'm thinking of going to Japan next year." That pattern alone will carry an enormous amount of ordinary conversation about your own life.

Conditional ば — hypotheticals. 時間があれば行きます (jikan ga areba ikimasu) — "If I have time, I'll go." Useful, but genuinely lower priority than everything above it.

The order to learn the forms for speaking

Textbooks order forms by grammatical tidiness. Speakers need them ordered by how many sentences each one unlocks per hour of study. That gives a different sequence:

1. ます — you can be polite about anything immediately. 2. て — clause linking plus six high-frequency patterns. 3. Past た — same sound change as て, so learn it the same week. 4. ない — you cannot have a real conversation without negation. 5. Potential — the form you need to describe yourself as a learner. 6. Volitional — invitations, and 〜ようと思います for plans.

Then, later and in no urgent order: conditional ば, たら, imperative, and the passive and causative — which are common enough to matter but complicated enough to deserve their own guide.

The reason this order works is that forms 1 to 4 are what almost every sentence in ordinary conversation is made of. A learner who has ます, て, た and ない fully automatic can talk for an hour. A learner who has studied all nine forms but produces none of them fluently freezes in the first thirty seconds — which is a distinct problem with its own set of fixes.

Out loud conjugation drills

Reading a conjugation table builds recognition. Speaking builds production. They are separate skills, and only the second one shows up in conversation. These three drills take five minutes a day and are done out loud, not in your head — the mouth needs the practice more than the eyes do.

Drill one the conjugation chain

Pick one verb. Say six forms in a fixed order, at speed, without pausing to think. The fixed order is what makes it automatic.

飲む → 飲みます → 飲んで → 飲んだ → 飲まない → 飲める → 飲もう nomu → nomimasu → nonde → nonda → nomanai → nomeru → nomō

食べる → 食べます → 食べて → 食べた → 食べない → 食べられる → 食べよう taberu → tabemasu → tabete → tabeta → tabenai → taberareru → tabeyō

する → します → して → した → しない → できる → しよう suru → shimasu → shite → shita → shinai → dekiru → shiyō

来る → 来ます → 来て → 来た → 来ない → 来られる → 来よう kuru → kimasu → kite → kita → konai → korareru → koyō

Ten verbs a day, roughly four minutes. Add 書く, 話す, 買う, 待つ, 帰る, 見る once the first four are effortless.

Drill two the trap sort

Say the ます form of these ten verbs out loud, then check against the table above. Any hesitation means the verb is not learned yet.

帰る · 入る · 走る · 切る · 知る · 着る · 見る · 起きる · 寝る · 要る

Six are Group 1 and four are Group 2. If you got 帰ます or 切ます, you have found exactly what this drill is for.

Drill three one verb five sentences

Take a single verb and produce five real sentences about your own life, one per form. Real content matters — invented textbook sentences do not transfer.

毎日コーヒーを飲みます (mainichi kōhī o nomimasu) — "I drink coffee every day." 今、コーヒーを飲んでいます (ima, kōhī o nonde imasu) — "I'm drinking coffee now." 昨日、お酒を飲みました (kinō, osake o nomimashita) — "I drank alcohol yesterday." お酒はあまり飲みません (osake wa amari nomimasen) — "I don't drink much alcohol." ビールが飲めません (bīru ga nomemasen) — "I can't drink beer."

This drill is where conjugation stops being a table and starts being speech. It is also where errors surface — and an error you never notice never gets fixed, which is the case for the correction feedback loop that most self-study lacks entirely.

For hearing these forms in real sentences rather than drill lists, NHK News Web Easy publishes short news items with furigana, and the Japan Foundation produces learner material organised around what you can do with the language rather than around grammar sequence.

Suru compounds are the biggest shortcut in Japanese

する looks like a small irregular verb. It is actually the most productive grammatical machine in the language, and understanding this is worth more than memorising fifty verbs.

Any noun of action plus する becomes a verb. Not some nouns — essentially any of them, including words borrowed into Japanese last year.

• 勉強 (study) → 勉強する benkyō suru — to study
• 電話 (telephone) → 電話する denwa suru — to phone
• 練習 (practice) → 練習する renshū suru — to practise
• 説明 (explanation) → 説明する setsumei suru — to explain
• 予約 (reservation) → 予約する yoyaku suru — to reserve
• 運転 (driving) → 運転する unten suru — to drive
• 準備 (preparation) → 準備する junbi suru — to prepare
• 心配 (worry) → 心配する shinpai suru — to worry

And the same applies to loanwords, which is how Japanese absorbs new vocabulary faster than any dictionary can track: コピーする (kopī suru, to copy), チェックする (chekku suru, to check), キャンセルする (kyanseru suru, to cancel), ダウンロードする (daunrōdo suru, to download).

Why this matters for conjugation: every one of these verbs conjugates identically, because they all end in する. Learn する once and you have conjugated thousands of verbs.

• 勉強します benkyō shimasu · 勉強して benkyō shite · 勉強した benkyō shita · 勉強しない benkyō shinai · 勉強できる benkyō dekiru · 勉強しよう benkyō shiyō

Note the potential form: 勉強できる, not 勉強しられる. The irregular potential of する carries straight through into every compound, and it is the form you will use most.

One structural detail worth knowing: many する compounds can also be split with を — 勉強をする is as correct as 勉強する, with a slightly heavier, more deliberate feel. And in formal or business Japanese the compounds multiply further, since a large share of workplace vocabulary is exactly this pattern: 確認する (confirm), 対応する (handle), 検討する (consider), 報告する (report). Anyone heading into a Japanese workplace will meet them within the first hour, which is why they dominate our business Japanese conversation guide.

The bottom line

Three groups. Group 1 shifts the final vowel, Group 2 drops る, Group 3 is する and 来る. Sorting a verb into its group is the entire difficulty of Japanese conjugation, and the ru-verb traps — 帰る, 入る, 走る, 切る, 知る and their handful of companions — are the entire difficulty of sorting.

After that, the forms are mechanical and there are fewer of them than in most European languages. No agreement with the subject, no gender, no plural, one exception in the て-form table, two irregular verbs in the whole language.

The part that is not mechanical is speed. Knowing that 飲む becomes 飲んで is not the same skill as producing 飲んで while you are also choosing a particle, retrieving a noun, and trying to remember whether you have already said this. That gap only closes with output, which is why the drills above are worth more than a second reading of the tables.

Which is where SoraSpeak fits: you speak out loud in the browser, an AI Japanese teacher replies in natural Japanese with subtitles, and every conjugation, particle and politeness error comes back explained in simple English while you still remember what you were trying to say. Being honest about it: it is a new product, it complements practice with real people rather than replacing it, and no software fully substitutes for a good teacher. For the specific job of saying a lot of conjugated sentences and finding out which ones were wrong, it is lower-friction than anything else available.

If you are also working towards a certificate, the official JLPT site lists which forms appear at each level — though note it tests reading and listening only, so a conjugation you can recognise on paper is not the same as one you can produce out loud. Next up, the two form families most learners skip entirely: the passive and causative.

🗣️ Three groups, nine forms, said out loud and corrected the same minute you get them wrong. Practise speaking Japanese free →

SoraSpeak is an independent Japanese speaking-practice app. Example sentences are written for learners; natural usage can vary by context and region.

FAQ