Oral Automatisms
A track of P45 · Speaking & Pronunciation.
The production face of the written automatism drills: hear a dialogue turn, say the reply aloud before you can read it, and get a word-by-word pronunciation score against a native model. Choosing the right sentence from four proves you know the rule; saying it at speaking speed proves the habit.
You can pick the correct sentence out of four. You do it every time, and you still say on a la appelée on a Tuesday afternoon in a shop. Both facts are true at once, and they are not in contradiction: choosing from four options is a knowledge test, and the errors that survive B2 are not knowledge failures. They are timing failures. The rule was available; it just arrived after the sentence had already left.
This track takes the automatism drills you already met in written form — the clitic chain and its elisions, the être reflex for pronominals and movement verbs, il y a, and the pronoun that has to carry a gender forward from a noun already gone — and removes the four options. You hear the question. You say the reply, out loud, before you can read anything. Only then does the model text appear, together with a word-by-word pronunciation score and the native recording to compare against.
The hidden reply is the whole design. If the sentence is on the screen while you speak, you are reading aloud, and reading aloud is a different exercise with its own track. What is being measured here is whether the form assembles itself in the half second before you open your mouth — which is exactly the half second in which the error happens.
Nothing new is taught. Every turn is a sentence the written drill already teaches, so a reply you fail here points at a rung you can go back to. The two surfaces count separately: saying four turns aloud does not consume four written items, because knowing them and being able to say them are different states and each needs its own evidence.