Who actually interacted? Scoring a recorded small-group discussion
Your Sec 5 enriched group of 26 has just finished a 12-minute small-group discussion on whether cellphones should be banned in class. Students read a 200-word prompt ten minutes earlier, and you recorded each group on a tablet. The rubric has three criteria: interaction and negotiation of meaning (40%), fluency and pronunciation (30%), accuracy (30%). Listening back to Group A, you note the following. Maude speaks for about seven minutes, opens every thread and interrupts twice. Julien gives four short turns, all grammatically clean, read aloud from his notes. Sofia asks three genuine questions, reformulates one of Julien's ideas, and repairs her own verb aloud ('ils ont… ils sont allés'). Noah says nothing for five minutes, then delivers one long, polished turn that sounds rehearsed. The program of study states that oral interaction is assessed on the student's ability to interact, not merely to produce sentences. You have 40 minutes to finalize marks before the report card.
1. Assign a mark out of 5 on the interaction criterion to Maude, Julien and Sofia, citing the evidence you used for each.
Read the worked answer
Maude - 3/5. She does initiate and sustain, which is real interactional behaviour. But the criterion measures negotiation among participants, and speaking for seven of twelve minutes means the others had little space to negotiate. The two interruptions are direct evidence against the criterion, not a personality quirk: turn-taking is part of the construct. She could reach 4 if the next recording shows her inviting a turn or building on someone else's idea.
Julien - 2/5. His turns are accurate but read from notes, so they demonstrate production, not interaction. Four short turns with no questions, no reformulation and no response to a partner's point give you almost no evidence of negotiation of meaning. Do not let the clean grammar pull the interaction mark up; accuracy is a separate 30%.
Sofia - 4.5/5. She asks three questions, reformulates, and repairs her own verb publicly. Self-repair inside talk is interactional competence: she monitors and fixes meaning in real time. The marker should be told explicitly not to also reward that repair under accuracy, or the same behaviour is paid twice.
2. What do you do about Noah - score the turn you have, re-record him, or something else? Justify your decision on validity and fairness grounds.
Read the worked answer
A single rehearsed monologue provides no evidence for a criterion about interaction, so scoring it as if it were interaction would be invalid. Giving him a second, friendlier task while everyone else was recorded once raises an equity problem if the second performance simply replaces the first, because the conditions of evidence collection would differ.
The defensible middle route is a supplementary sample: a 3-minute paired decision task with one partner (for example, agreeing on three phone rules), scored only on the interaction criterion, with the original recording kept for fluency and accuracy. This adds a missing data point rather than erasing a weak one, and it is documented as an accommodation available to any student whose group configuration prevented them from interacting.
Also hold one interpretation in reserve: five minutes of silence may be anxiety or group dynamics, not low competence. One more sample distinguishes those before you write a mark that follows him into Sec 5 records.
3. A student with strong grammar argues that 40% for interaction is unfair because 'grammar is the real French.' Write the reasoning you would give, and propose one change to the task that would make the interaction criterion easier to evidence.
Read the worked answer
The 40% is not a judgment that grammar matters less; it reflects where communication actually breaks down. When a learner cannot ask for clarification, hold a turn, or check that a partner understood, the exchange collapses regardless of how well the verbs are conjugated. Accuracy keeps 30%, so grammar is still a substantial share - it is simply measured where it belongs rather than being allowed to stand in for the whole competency.
There is also an acquisition argument worth stating plainly: risk-taking in interaction is what generates the input and feedback that move a learner forward. A student who speaks only when certain produces fewer errors and learns more slowly.
Task change: give each participant a role card with information the others do not have, and require the group to reach a written consensus in ten minutes with a designated reporter. Information gaps force questions and clarification, which produces audible evidence of negotiation instead of leaving it to chance. A time limit with rotating roles also reduces the dominance problem visible in Maude's seven minutes.