What a drill actually trains
A drill removes choice. The pattern is given, the vocabulary is usually given, and what remains is producing the form accurately and quickly. That narrowness is the point: it gets the repetitions that conversation cannot supply, since a real conversation might contain one instance of the pattern you are working on.
What a drill does not train is selection. Having produced the conditional fifty times when told to, you have not practised noticing that a conditional is what this situation needs, and that noticing is most of the difficulty in speech.
What conversation actually trains
Conversation trains everything at once and nothing intensively. You get retrieval under pressure, selection, listening, repair and timing, all of which are the real skills, but you get one or two instances of each specific form.
This is why conversation alone leaves errors in place. Fifty conversations might contain fifty instances of your particle problem, each too brief to fix, and no occasion where you slow down enough to produce the right version deliberately.
| Skill | Drills | Conversation |
|---|---|---|
| Accuracy on one form | Strong | Weak |
| Speed on one form | Strong | Moderate |
| Choosing the right form | None | Strong |
| Listening and repair | None | Strong |
| Stamina | None | Strong |
| Pronunciation | Strong if scored | Weak, unmonitored |
Read the table as an argument for both. Every strong entry in one column is a gap in the other.
A split that works
One part drilling to two parts conversation is a reasonable default for anyone past the first few months. In a forty-five minute session that is fifteen minutes of targeted drilling and thirty of speaking, with the drill first so it is fresh when the conversation starts.
The order matters more than the ratio. Drilling after a conversation, on the errors that conversation revealed, is the highest-value version, and drilling immediately before the next conversation is what makes the pattern available when it is needed.
- Ten to fifteen minutes: one pattern or one sound, spoken, at half speed then full.
- Twenty-five to thirty minutes: unscripted conversation, no notes.
- Ten minutes: review the corrections and pick tomorrow's drill.
- Once a week: skip the drill entirely and just talk for the full time.
Making the drill transfer
The link between drill and conversation has to be built deliberately, because it does not form on its own. The technique is simple: decide before the conversation that you will use the drilled pattern at least three times, and count.
This feels artificial and it works. Announcing the intention to your practice partner works even better, since the situations that call for the pattern then get created for you.
| Japanese | Romaji | Meaning |
|---|---|---|
| 今日は「たら」を使ってみます。 | Kyō wa tara o tsukatte mimasu. | Today I will try to use tara. |
| この文型の練習をしたいです。 | Kono bunkei no renshū o shitai desu. | I want to practise this pattern. |
| 使えていなかったら、教えてください。 | Tsukaete inakattara, oshiete kudasai. | If I am not using it, please tell me. |
| わざと難しい言い方をしてみます。 | Waza to muzukashii iikata o shite mimasu. | I will deliberately try a harder phrasing. |
| 今の、自然でしたか。 | Ima no, shizen deshita ka. | Was that natural just now? |
| もう一度、別の言い方で言います。 | Mō ichido, betsu no iikata de iimasu. | I will say it again a different way. |
Letting conversation set the drill list
The most common mistake is drilling what a textbook is covering rather than what your speech is missing. Textbook order is designed for coverage; your errors are the actual curriculum, and they are far shorter.
Take the drill list from your last conversation, keep it to two items, and drop anything that has not appeared in a real conversation this month. This keeps the drilling small enough to actually do.
- One grammar pattern that failed, or came out wrong twice.
- One sound that a score or a listener flagged.
- Nothing else, however tempting the textbook chapter looks.
Running the loop in one place
The loop is: converse, review, drill, converse again with intent. Its weak point is the review, because reconstructing what went wrong from memory is unreliable and tedious, which is why most learners skip it and go straight back to conversing.
In Unihongo the review step is already done for you. The session report after each conversation shows what you said, the corrected version and what to work on, so the drill list writes itself. Sound problems go to the pronunciation lab, which scores individual lines against a native model; vocabulary goes to flashcards, which bring the words back before the next session.
Then the next conversation in the immersive 3D classroom is the test, spoken and unscripted, rather than another round of typing where you can compose at leisure. That is the whole loop, and it is the structure that makes drilling worth its time.
If you can only keep one part of the loop, keep the review. Conversation without review is the expensive way to stay where you are.