Two ways learners get through a conversation
The phrase-based speaker has a stock of complete sentences, retrieved whole. They sound remarkably good in the situations they prepared for, because a memorised sentence carries native word order, correct particles and appropriate register with no processing cost at all.
The rule-based speaker builds from grammar. They are slower and make more errors, but they can say things nobody taught them, which means they can survive a conversation that goes somewhere new.
Neither is a personality type. Most learners drift into one because of how they studied, and the fix is to notice which one you default to and deliberately practise the other.
What memorised phrases do well
Some Japanese is genuinely fixed, and trying to build it from grammar produces sentences that are technically correct and clearly wrong. Set expressions for greetings, apologies, entering and leaving, and receiving food are conventional in a way that rules do not capture.
These are worth memorising exactly, with their rhythm, because they are also the phrases said most often and heard most closely.
| Japanese | Romaji | Meaning |
|---|---|---|
| お疲れさまです。 | Otsukaresama desu. | Thank you for your work. |
| お世話になっております。 | Osewa ni natte orimasu. | Thank you for your continued support. |
| 恐れ入りますが、 | Osoreirimasu ga, | I am sorry to trouble you, but |
| よろしくお願いいたします。 | Yoroshiku onegai itashimasu. | I look forward to working with you. |
| 失礼いたします。 | Shitsurei itashimasu. | Excuse me. (on leaving) |
| いただきます。 | Itadakimasu. | Thank you for the meal. (before eating) |
Do not analyse these. They are conventions, and using them at the right moment matters far more than understanding their internal grammar.
Where a memorised script breaks
A script fails at the second turn. The memorised opening works, the other person answers in a way you did not rehearse, and now you are in a conversation you started but cannot continue. This is the specific failure mode of phrase-based learning, and it is why phrasebook preparation feels so much better in the hotel room than in the shop.
It also fails on register. A phrase learned in one setting carries the politeness level of that setting, and produced in a different one it sounds either stiff or rude. A phrase you cannot vary is a phrase you cannot adjust.
- The reply falls outside the script, and there is no second line.
- The situation is slightly different, and the phrase does not quite fit.
- The register is wrong for this listener, and you cannot soften it.
- You need the negative, the past, or a question form, and only have the one form.
Turning a phrase into a pattern
The conversion is simple and it is the highest-value thing you can do with a phrasebook. Take the memorised sentence, find the part that changes, and practise substituting it until the substitution costs nothing. One phrase becomes a template that covers dozens of situations.
Do it out loud, with at least five substitutions, and include the awkward ones. The variation you never practise is the one the conversation will ask for.
| Memorised phrase | The variable | Pattern |
|---|---|---|
| これをください | the object | ___ をください |
| トイレはどこですか | the place | ___ はどこですか |
| 写真を撮ってもいいですか | the action | ___ てもいいですか |
| 行きたいです | the verb | ___ たいです |
| 高いと思います | the adjective | ___ と思います |
| すみません、遅れます | the situation | すみません、___ |
Five patterns with confident substitution will carry more conversations than fifty memorised sentences, because they cover situations nobody anticipated for you.
Improvising with the grammar you already have
Improvisation does not require more grammar. It requires willingness to say the simpler thing, which learners resist because the simpler thing feels like a concession. It is not: a listener would far rather hear a plain sentence now than a sophisticated one in ten seconds.
The technique is to lower the ambition of the sentence rather than the ambition of the idea. Split it into two simple sentences, drop the subordinate clause, or describe rather than name.
| Japanese | Romaji | Meaning |
|---|---|---|
| 名前は分からないんですが、丸くて、赤いものです。 | Namae wa wakaranai n desu ga, marukute, akai mono desu. | I do not know the name, but it is round and red. |
| 簡単に言うと、こういう仕事です。 | Kantan ni iu to, kō iu shigoto desu. | Put simply, it is this kind of job. |
| 二つに分けて説明します。 | Futatsu ni wakete setsumei shimasu. | I will explain it in two parts. |
| 日本語でなんて言うか分かりません。 | Nihongo de nante iu ka wakarimasen. | I do not know how to say it in Japanese. |
| 同じような意味の言葉でいいですか。 | Onaji yō na imi no kotoba de ii desu ka. | May I use a word with a similar meaning? |
| つまり、こういう感じです。 | Tsumari, kō iu kanji desu. | So, it is something like this. |
Practising the handover
In real speech the two modes alternate constantly, often inside one sentence: a memorised opening, an improvised middle, a memorised close. The skill worth training is the handover, the moment where the script runs out and you have to continue without it.
That moment only appears when the conversation is unpredictable. Speaking with an AI Sensei in Unihongo's immersive 3D classroom produces it reliably, because the follow-up depends on what you actually said rather than on a fixed branch, and because answering aloud gives you no time to compose the way a text chatbot does.
Afterwards the session report shows the handovers that failed, and the phrases involved can be saved as flashcards in their pattern form rather than as fixed sentences, so the next round of review trains the variable part.
A good sign of progress is hearing yourself modify a memorised phrase mid-sentence. That is the two systems working together rather than one covering for the other.