Rich Language, Wrong Reader: Scoring a Register Failure

Neither is strong; they fail on different axes. Indicative scoring:

• Student A: task completion ~75% of that strand - the message is complete and the critical information is present, but 190 words exceeds the limit (a real constraint in an exam task, so it costs something). Language control ~90%. Register ~25% - « Salut » plus tu to an adult host mother you have never met, plus « Bizoux » as a sign-off, is a genuine failure to manage the relationship the task creates. Total ≈ 60%.
• Student B: task completion ~35% - the single piece of information the email exists to convey is missing. The whole point. Language control ~85% (short, so fewer opportunities). Register ~100%. Total ≈ 63%.

What the rubric forces you to decide: whether conventional politeness and communicative success are the same thing. They are not. B has learned a formula and deployed it as a substitute for content; A has content and no audience awareness.

The defensible call: B's omission is the more serious failure, because an email that does not state the arrival time cannot do its job, whereas A's email would still get her collected from the station - the host mother would wince, then reply. If your rubric produces B ≥ A, reweight: a missing obligatory content point should cap the task-completion strand hard, whatever the tone.

What I would say (written, plus two minutes face to face):

1. Name the strength first, concretely. "Your subjonctifs and your future perfect are correct and they're doing real work - « il faudra que je sache » is exactly right."
2. Frame register as a decision about the reader, not a rule. "Who is Martine to you in this scenario? An adult you've never met, who is doing you a favour. In French, vous is the default until she proposes tu - and she hasn't, because she hasn't written back. So you were the one deciding the relationship, in the first line, before saying anything."
3. Show, don't tell. Have her rewrite only the opening and closing lines, then compare: « Bonjour Madame Delaunay, » … « Bien cordialement, » against her original. Two lines, thirty seconds, high retention.
4. Handle « Bizoux » precisely. It's fine between friends; it is not informal-French-is-bad. The error is calibration, not vulgarity.
5. Name « réponds-moi vite ». An imperative to someone doing you a favour reads as an order. « Pourriez-vous me confirmer… ? » does the same work.

What I would not say: "French people are formal." That is a stereotype, it is false at the level of actual French social life, and it teaches nothing transferable. The rule to teach is the other person sets the register, not the French are stiff.

Change the prompt's framing, not the rubric.

Add one line to the task: « Martine Delaunay, la mère de ta famille d'accueil, t'a écrit une fois ; elle t'a vouvoyée. » That single sentence converts the register strand from a guess into a judgement: the student now has evidence about the relationship and must decide what to do with it. Right now the prompt withholds exactly the information the 20% strand is testing, which makes that 20% partly a lottery on whether the student happens to know French host-family convention.

Why not the other candidates:

• Reweighting task completion higher fixes this pair of scripts but not the underlying design flaw - the student still has nothing to reason from.
• Requiring a formal salutation in the instructions removes the test. You would be measuring compliance, not appropriateness.
• Dropping the word limit removes a real constraint that distinguishes A from a strong response; keep it, and enforce it.

Secondary fix: break the task completion strand into a checklist of obligatory content points (arrival time, origin, date, request for confirmation) worth a fixed value each, so that a missing critical point cannot be averaged away by elegant prose. That is what made B's score indecent, and it is a five-minute rubric edit.