This is the most-consulted table in the whole guide, and for good reason: one map governs the te-form, the plain past た, and every te-based auxiliary (〜ている, 〜てください, 〜てから, 〜たら, 〜たり…). Learn this chart once and a dozen grammar points stop being separate memorization jobs. The changes it describes are called 音便（おんびん, "euphonic sound changes"）: as the connector て/た historically glued onto the verb stem, the stem's final consonant softened to make the join easier to pronounce. Only the 五段（ごだん） verbs undergo them; 一段 verbs and the irregulars are trivial. Below is the reference row, the full chart, and the two errors it exists to prevent.
The reference row: 書く → 書いて／書いた
Anchor everything to 書く（かく, "to write"）. Its te-form 書いて and past 書いた show the pattern that a beginner meets first, and the mechanism — く softening to the vowel い — is the template for reading the rest of the chart.
ここに名前を書いてください。
koko ni namae o kaite kudasai
Please write your name here.
昨日、母に手紙を書いた。
kinō, haha ni tegami o kaita
Yesterday I wrote a letter to my mother.
The master chart — every ending, te and た together
| Dictionary ending | 音便 type | te-form | Past た-form | Model verb |
|---|---|---|---|---|
| -う | 促音便 (っ) | 会って (atte) | 会った (atta) | 会う (au) — meet |
| -つ | 待って (matte) | 待った (matta) | 待つ (matsu) — wait | |
| -る (五段) | 取って (totte) | 取った (totta) | 取る (toru) — take | |
| -ぬ | 撥音便 (ん, voiced → で／だ) | 死んで (shinde) | 死んだ (shinda) | 死ぬ (shinu) — die |
| -ぶ | 遊んで (asonde) | 遊んだ (asonda) | 遊ぶ (asobu) — play | |
| -む | 読んで (yonde) | 読んだ (yonda) | 読む (yomu) — read | |
| -く | イ音便 (い) | 書いて (kaite) | 書いた (kaita) | 書く (kaku) — write |
| -ぐ | イ音便 (い, voiced → で／だ) | 泳いで (oyoide) | 泳いだ (oyoida) | 泳ぐ (oyogu) — swim |
| -す | no 音便 (し-stem) | 話して (hanashite) | 話した (hanashita) | 話す (hanasu) — speak |
| -る (一段) | drop る | 食べて (tabete) | 食べた (tabeta) | 食べる (taberu) — eat |
| する | irregular | して (shite) | した (shita) | する (suru) — do |
| 来る | irregular | 来て (kite) | 来た (kita) | 来る (kuru) — come |
| 行く | exception (っ) | 行って (itte) | 行った (itta) | 行く (iku) — go |
The three 音便 groups, explained
The whole chart is really just three softening patterns plus two "no-change" rows. Understanding why each group behaves as it does lets you reconstruct any cell instead of memorizing forty forms.
促音便（そくおんびん, "doubling sound-change"）— う・つ・る → って／った. The final consonant collapses into a small っ, doubling the following t. 会う → 会って, 待つ → 待って, 取る → 取って. These three endings all had a stem consonant that assimilated cleanly into the connector.
駅前で三十分も待った。
ekimae de sanjuppun mo matta
I waited a full thirty minutes in front of the station.
撥音便（はつおんびん, "nasal sound-change"）— ぬ・ぶ・む → んで／んだ. The stem nasalizes to ん, and — critically — the nasal voices the connector: て becomes で, た becomes だ. 死ぬ → 死んで, 遊ぶ → 遊んで, 読む → 読んで. There is no ×死んて in the language.
週末、友達と海で遊んで、たくさん写真を撮った。
shūmatsu, tomodachi to umi de asonde, takusan shashin o totta
At the weekend I hung out with friends at the beach and took lots of photos.
イ音便（イおんびん, "い sound-change"）— く → いて／いた, ぐ → いで／いだ. The final consonant softens to the vowel い. Voiceless く keeps the plain connector (書いて); its voiced twin ぐ voices the connector to で/だ (泳いで／泳いだ), for the same reason the nasal group does.
そのニュースを聞いて、本当に驚いた。
sono nyūsu o kiite, hontō ni odoroita
When I heard that news, I was truly astonished.
No 音便 — す → して／した; 一段 → て／た. The -す verbs simply take their し-stem plus て (話して) with no softening at all, and 一段 verbs just drop る and add て (食べて). These are the "easy" rows.
先生とじっくり話して、少し気持ちが楽になった。
sensei to jikkuri hanashite, sukoshi kimochi ga raku ni natta
I talked things over with my teacher and felt a bit lighter.
The two irregulars
する → して／した and 来る → 来て／来た. Note the reading of 来て: it is kite (the き-row), not kote or kute. The kanji 来 hides a vowel that shifts across the paradigm (来る kuru, 来て kite, 来ない konai), so anchor it by sound, not by the character.
ちょっとこっちに来て、これ見て。
chotto kocchi ni kite, kore mite
Come here a sec and look at this.
The one exception in the entire system: 行く → 行って
行く（いく, "to go"）ends in く and should, by the イ音便 rule, give ×行いて. It does not. 行く takes っ instead: 行って／行った, patterning like a つ- or る-verb. This is the single high-frequency irregular in the te/ta system — and because "to go" is one of the commonest verbs in the language, you will produce it constantly. Every other cell of 行く is regular (行かない, 行きます, 行ける); only the te/ta break. Do not over-generalize the exception: every other -く verb is a well-behaved 書いて-type (歩く → 歩いて, 聞く → 聞いて). The full account lives on 行く: the te-form exception.
ちょっとコンビニに行って、すぐ戻る。
chotto konbini ni itte, sugu modoru
I'll just pop to the convenience store and be right back.
Why English gives you no intuition here
English forms its past by changing the whole word as a unit ("write / wrote," "read / read"), and its connective ("write and…") never reshapes the verb at all. Japanese does the opposite: the ending itself softens so that the connector te/ta can attach smoothly, and that softening is systematic — three sound-laws, not a list of exceptions. There is genuinely nothing to reconstruct from English; instead, internalize the three 音便 groups above, and the te-form, the past, and every 〜て auxiliary become one skill. For a memory hook to drill the groups, see the te-form mnemonics page.
Common mistakes
❌ 昨日、本を読んて寝た。
Wrong — the ぬ・ぶ・む group voices to で／だ: 読む → 読んで／読んだ, never ×読んて.
✅ 昨日、本を読んで寝た。
kinō, hon o yonde neta
Last night I read a book and went to sleep.
❌ 先生と話って、安心した。
Wrong — -す never doubles to って; it takes して (no 音便): 話す → 話して.
✅ 先生と話して、安心した。
sensei to hanashite, anshin shita
I talked with my teacher and felt reassured.
❌ 海で二時間も泳いて、疲れた。
Wrong — -ぐ is voiced: 泳ぐ → 泳いで, matching the voicing rule.
✅ 海で二時間も泳いで、疲れた。
umi de nijikan mo oyoide, tsukareta
I swam for two hours in the sea and got tired.
❌ 図書館に行いて、本を借りた。
Wrong — 行く is the exception: 行って, not the regular ×行いて.
✅ 図書館に行って、本を借りた。
toshokan ni itte, hon o karita
I went to the library and borrowed a book.
❌ ご飯を食べって、出かけた。
Wrong — 一段 verbs take no 音便; they just drop る: 食べる → 食べて, never ×食べって.
✅ ご飯を食べて、出かけた。
gohan o tabete, dekaketa
I ate and then went out.
Key takeaways
- One chart governs te, past た, and every 〜て auxiliary — learn it once.
- Only 五段 verbs undergo 音便; there are exactly three groups: う・つ・る → って／った, ぬ・ぶ・む → んで／んだ, く → いて／いた and ぐ → いで／いだ.
- す takes no 音便 (話して／話した); 一段 verbs just drop る (食べて／食べた).
- The voiced connector で／だ appears only after the ん-group and ぐ — that single boundary prevents ×読んて and ×泳いた.
- Irregulars する → して, 来る → 来て（きて）; the lone exception is 行く → 行って, not ×行いて.
Now practice Japanese
Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.
Start learning Japanese→Related Topics
- Past た-Form: Conjugation TableN5 — The plain past た across every class — built exactly like the te-form but with た／だ, so the same 音便 map (including the voicing) applies throughout.
- 行く: The te-form ExceptionN4 — 行く（いく, to go）is a perfectly regular 五段 -く verb in every cell except one — its te-form and past are the 促音便 forms 行って／行った, never the ×行いて that 書く predicts.
- All Forms, All Classes: Master ChartN4 — The one-sheet everything reference — every major verb form (dictionary through causative-passive, volitional, conditional, imperative) down the side and 書く・食べる・する・来る across the top, so you can verify any form without hunting across pages.
- The て-form: Japanese's Universal ConnectorN4 — Why the tenseless, politeness-free て-form is the single most productive conjugation in Japanese — the hinge that feeds requests, progressives, sequence, permission, and dozens more constructions.
- The te-form Song: All Rules on One PageN4 — The complete te-form system on a single page, built around the classic learner mnemonic — う・つ・る→って, む・ぬ・ぶ→んで, く→いて, ぐ→いで, す→して, plus ichidan and the two irregulars.