Textbook audio pronounces every auxiliary in full: 食べている, 食べてしまう, しなければならない. Nobody at a real dinner table does. Spoken Japanese compresses its most common grammar into a small set of contractions, and until your ear knows them, casual conversation stays a blur no matter how much vocabulary you know.
The good news is the same as for English gonna and wanna: the reductions are systematic, not random. There are a couple dozen mappings worth owning, they apply everywhere, and each one is a full grammatical form you have probably already studied — just faster.
The contractions that carry casual speech
| Textbook form | What you actually hear | What it does |
|---|---|---|
| 〜ている / 〜でいる | 〜てる / 〜でる | Progressive and result states: 食べてる, 住んでる |
| 〜てしまう / 〜でしまう | 〜ちゃう / 〜じゃう | Done completely, or with regret: 食べちゃった |
| 〜ておく / 〜でおく | 〜とく / 〜どく | Done in advance: 買っとく |
| 〜なければならない | 〜なきゃ / 〜なくちゃ | Obligation, clipped: 行かなきゃ |
| 〜ければ (conditionals) | 〜けりゃ / 〜りゃ | 安ければ → 安けりゃ |
| 〜のだ | 〜んだ | Explanatory tail: そうなんだ |
| 〜では | 〜じゃ | これじゃ, 雨じゃ |
| と言っている / と言う | 〜つって / 〜って | Quotation, compressed: 行くつってた? |
Two things to notice. First, the compression always eats the same sound: the て-row auxiliary loses its vowel, and the trailing い of いる goes first. Learn the rule once and you can decompose forms the table does not list. Second, none of these are recent inventions or youth slang — your grandmother says てる. They are simply the default register of spoken Japanese, the way going to becomes gonna in every English accent on earth.
Hearing them in the wild
Casual register is more than contracted auxiliaries — it is a whole posture, with fillers, dropped politeness, and sentence-final particles doing the emotional work. A breakfast-diary vlog from the corpus, narrating a rushed morning:
サッと食べるね。やっぱり朝子供たちもバタバタだし
“You can eat it quickly — mornings with the kids are hectic, too.”
No です, no ます. The ね invites agreement, the し chains the reason on without a connective, and the verb stands in bare dictionary form. Or the verdict line that closes meals across food vlogs:
ごちそうさまでした。めっちゃ美味しかった
“Thank you for the meal — it was really delicious.”
めっちゃ is the casual intensifier that replaces とても, and the opinion lands in plain past tense, 美味しかった, because narrating to a camera is talking to a friend, not to a teacher. If your ear trained only on polite audio, these lines feel “fast” — but read them back and every word is textbook-simple. The speed was register, not pace.
The polite-first training order
Learners sometimes ask whether to skip polite forms and drill casual speech from the start, since it is “the real language”. Don’t. Polite です/ます speech is slower, evenly paced, and grammatically complete — it is the right ear food while the foundation sets, and it is also the register you will personally need at every counter and restaurant for years. The case for polite-first training is made in detail in the food-vlog guide.
The efficient sequence is: own the full form first, then train its contraction as a reduction. なきゃ decodes instantly if なければならない is already automatic; it is opaque otherwise. This is why contractions are an N4 project, not an N5 one — they compress N4 grammar, and by N4 you have the raw material.
How to drill them
Contractions yield to the same routine as any listening feature, with one twist — you count:
- Play a casual-speech video with word-level subtitles. When a contracted form blurs past — てる, ちゃう, なきゃ — tap it and replay the exact second. Hear where the て-row lost its vowel. Each word page on this site also lists real corpus sentences, so 食べる, heard in 88 of the 131 current videos, can be drilled in its inflected wild forms directly.
- Save the contracted form, not just the dictionary form. The card that reviews 食べちゃう with its original audio teaches your ear the sound it will actually meet; a card that reviews only 食べる leaves the contraction unheard. Let FSRS spaced repetition resurface them at the forgetting edge — the algorithm discussion in FSRS vs SM-2 explains why that timing cements them.
- Count hits per video. Second viewing, keep a tally: every てる, every ちゃう, every って. Most videos serve a dozen before the two-minute mark, and counting converts an undifferentiated blur into a finite, owned list — the moment contractions stop being “fast speech” and start being vocabulary you happen to know.
Casual dialogue lives across the whole corpus — the N4 daily-life collection and N4 entertainment collection are full of narrators talking to the camera the way people actually talk. A week of counting contractions in those, and polite-form audio will start sounding conspicuously slow. That is the sound of your ear moving up a register.