Corrections without review change nothing
In the moment a correction feels like learning. You hear the right version, it makes sense, and the sense of understanding is convincing. But understanding a correction and being able to produce the form under pressure are different achievements, and only the second one changes your speech.
What is missing is a second encounter. A correction seen once behaves like any other single exposure: mostly forgotten within a day. A correction seen again the same evening, and again two days later in a context that requires producing it, behaves like something you have learned.
This is why the review routine matters more than the quality of the feedback. Excellent corrections that are never revisited lose to mediocre corrections that are.
Sorting corrections into three piles
A single conversation can produce twenty corrections, which is far too many to act on. Sorting them takes two minutes and makes the rest of the review possible, because it reduces twenty items to two or three.
Sort by what the error costs, not by how embarrassing it felt. Errors that block understanding come first, errors that repeat come second, and single occurrences of something you have not studied go on a list for later.
| Pile | What goes in it | Action |
|---|---|---|
| Fix now | Blocks meaning, or repeated | Drill and flashcard |
| Note | Unfamiliar grammar, once | Add to a list, study later |
| Ignore | Style, rare words, slips | Nothing |
| Sound | Length, pitch, small tsu | Pronunciation lab drill |
The ignore pile is important. Giving yourself permission to drop most corrections is what makes it possible to act on the few that matter.
A ten-minute review that fits after a session
The routine is fixed so that it requires no decisions when you are tired. Sort, say aloud, write two cards, and note one thing to do differently next time. It takes ten minutes and it is the difference between practice that accumulates and practice that repeats.
The saying aloud step is not optional. Reading a correction silently rehearses recognition; saying the corrected sentence five times rehearses production, which is the skill that failed.
- Two minutes: sort the corrections into the three piles.
- Four minutes: say each fix-now correction aloud five times in a full sentence.
- Two minutes: write two flashcards, no more.
- Two minutes: write one sentence about what to do differently next session.
Turning a correction into a flashcard
Most correction flashcards fail because they test the wrong thing. A card with the Japanese on the front and the English on the back tests recognition, which was never the problem: you understood the correction perfectly when you heard it.
The card that works puts the situation on the front and the Japanese on the back. It forces retrieval under something like the conditions of speech, and it re-creates the moment where the error occurred, which is what you actually need to change.
| Japanese | Romaji | Meaning |
|---|---|---|
| 先週の会議のことを話す | Senshū no kaigi no koto o hanasu | Front: talking about last week's meeting |
| 先週の会議で、その件を説明しました。 | Senshū no kaigi de, sono ken o setsumei shimashita. | Back: the corrected sentence |
| 店員に袋を断る | Ten'in ni fukuro o kotowaru | Front: refusing a bag at the till |
| 袋は大丈夫です。 | Fukuro wa daijōbu desu. | Back: the natural version |
| 理由を言って断る | Riyū o itte kotowaru | Front: declining with a reason |
| その日は用事があって、行けないんです。 | Sono hi wa yōji ga atte, ikenai n desu. | Back: the full sentence |
Re-testing the same mistake next session
The last step is the one almost nobody does: deliberately walking back into the situation that produced the error. Reviewing a correction proves you can produce the form when you are thinking about it. Re-entering the situation proves you can produce it when you are not.
Decide the re-test before the session starts, so the target is fixed. One error, one situation, and a note afterwards about whether it appeared again.
| Japanese | Romaji | Meaning |
|---|---|---|
| 前回の話の続きをしてもいいですか。 | Zenkai no hanashi no tsuzuki o shite mo ii desu ka. | May we continue the previous topic? |
| 同じ場面をもう一度練習したいです。 | Onaji bamen o mō ichido renshū shitai desu. | I want to practise the same scene again. |
| この言い方を直したいです。 | Kono iikata o naoshitai desu. | I want to correct this way of saying it. |
| 間違えたら、教えてください。 | Machigaetara, oshiete kudasai. | If I make a mistake, please tell me. |
| 特に助詞を見てください。 | Toku ni joshi o mite kudasai. | Please watch the particles especially. |
| 今日はこの表現を使ってみます。 | Kyō wa kono hyōgen o tsukatte mimasu. | Today I will try to use this expression. |
Naming your target at the start of a session also changes the feedback you get, because the corrections come back focused on the thing you are working on.
A weekly audit that catches the patterns
Individual corrections are noisy. Patterns only show up across a week, and they are where the real progress is: the same particle wrong in four different sessions, or every conditional coming out as tara regardless of meaning.
Once a week, read back the fix-now piles from every session together. Two or three recurring items will stand out immediately, and those become the next week's accuracy drills. Sound errors that keep reappearing go to the pronunciation lab, where they can be scored line by line rather than guessed at.
Because each Unihongo session leaves a report behind, the week's material is already collected: what you said, the corrected version, and what to work on, session by session. The audit is reading rather than reconstructing, which is the difference between a routine you keep and one you abandon after a fortnight.
Expect the recurring list to be short and to change slowly. Three items that disappear over a month is a good month.