Fourteen errors in ten minutes: what to correct, when, and what it does to the mark
During a ten-minute picture-story task, students narrate in the past tense. Listening to Kevin's recording, you log 14 errors. Nine involve the passé composé: 'j'ai allé,' 'les personnes que j'ai vu,' 'elle est parti,' and similar forms. Two are lexical: 'actuellement' used to mean 'currently,' and 'je suis excité' used to mean 'I'm excited.' Three are phonological: a missing liaison in 'ils ont,' an unstable 'eu' vowel, and misplaced stress in 'développement.' Kevin's story is coherent, he uses connectors correctly (d'abord, ensuite, finalement), and he self-corrects once. The term's oral mark weights fluency and message at 60% and accuracy at 40%. A colleague advises correcting everything on the spot so errors do not fossilise; your department head says fluency tasks should never be interrupted.
1. Which of Kevin's 14 errors do you treat during the task, which after, and which not at all? Justify by error type and by the purpose of the task.
Read the worked answer
During the task: only the meaning problem. 'Je suis excité' is a genuine risk of misinterpretation in French, and a quick recast ('tu veux dire que tu étais impatient?') costs three seconds and protects the story. A recast, not a correction: you restate the target, you do not stop the narration. Nothing else is worth interrupting for - this is a fluency task, and every interruption buys accuracy with fluency.
After the task: the passé composé. But split it, because these are two rules, not one. 'J'ai allé' is auxiliary choice; 'elle est parti' and 'les personnes que j'ai vu' are agreement. Nine instances in ten minutes is a system, not slips, and a system needs a focused lesson, not a red pen: a sorting activity with the être verbs plus one rule on agreement with a preceding direct object. Correcting all nine individually in the margin teaches nothing transferable.
Not as accuracy at all: 'actuellement' belongs in a vocabulary and register lesson on faux amis, and the phonology gets two minutes of modelling in a warm-up - liaison first, since it affects listening as well as production.
2. Given the 60/40 split, show a defensible calculation for the term's oral mark and say what would make you adjust it.
Read the worked answer
Fluency and message - 52/60. Coherent narrative, correct sequencing, good connectors, one self-repair, sustained over ten minutes. Not higher because the self-repair and some hesitation suggest monitoring is still effortful.
Accuracy - 22/40. At Sec 5 enriched, agreement in the passé composé with être verbs is expected in familiar narration, and nine failures in one task is frequent, so this sits below expectation but is not a collapse. Crucially, do not deduct again under fluency for the same self-repair. Total: 74/100.
Adjustments: this is one ten-minute sample, and a single task is thin evidence for a term mark. Collect a second sample - a short written narration covering the same structures - and let the two inform each other; if the written version shows the auxiliary errors disappearing under planning time, that indicates a processing constraint rather than missing knowledge.
If the next sample shows him avoiding the passé composé altogether, the profile has changed and the accuracy mark should reflect avoidance, which is a stronger signal than error.
3. State the conditions under which on-the-spot correction actually helps, and name one signal that it is backfiring for Kevin.
Read the worked answer
On-the-spot correction helps when three conditions hold. The learner must notice the target - unnoticed recasts do nothing. The structure must be within reach, which for Kevin means the auxiliary choice is treatable now while nuanced agreement may not be. And the task's purpose must allow it: interrupting a fluency monologue changes the task into an accuracy task.
Form matters too: recasts suit phonology and lexis, while prompts ('tu es allé…?') suit rule-based grammar because they push the learner to retrieve the form rather than hear it.
The signal that it is backfiring is avoidance. If the next recording shows shorter turns, a shift into the present tense, or the disappearance of passé composé narration, the corrections have taught him that the past tense is dangerous. Track turn length and tense use before and after any change in feedback policy; that comparison is the evidence, not your impression of the lesson.
Practical compromise with your colleague: keep a three-item error log per student and correct from it after the task, rather than correcting fourteen things mid-sentence.