Scoring a Simulated Conversation with Register Shifts
Herr Keller teaches German 3 at a suburban high school with 24 students, including five heritage speakers. For the interpersonal speaking benchmark, students role-play asking a Munich host family for permission to invite a friend to dinner. The department rubric rates task completion, comprehensibility, vocabulary, and language control on a 1–4 scale; the benchmark is 15% of the quarter grade. Maya, a strong student, says: 'Entschuldigung, kann ich ... äh ... mein Freund kommen? Er isst gern. Ja, und ... wie sagt man ... er hat keine Allergie. Ich denke, das ist gut. Können Sie ... kannst du ...?' She pauses for four seconds and self-corrects between Sie and du twice. She uses essen instead of zu Abend essen but makes the request understandable. A second rater assigns 3/4 overall; Keller wants 2/4 because of the register shifts. Department procedure calls for a third rater only when holistic scores differ by more than one point.
1. How should Keller and the second rater resolve the disagreement, and what holistic score is defensible?
Read the worked answer
Separate the rubric dimensions rather than letting one salient feature drive the score. On task completion, Maya asks permission, identifies the friend, and gives a reason; that is a 3 or 4. Comprehensibility stays high because a listener can act on the request despite pauses. Vocabulary is adequate but generic ('isst gern' rather than 'zu Abend essen'), so 2–3. Language control is the weak dimension: the Sie/du self-corrections show awareness without control, and 'wie sagt man' plus the four-second pause signals recall difficulty. A holistic 3/4 is defensible because the weakness is localized to register and formulaic chunks, not global breakdown. Keller's 2/4 overweights language control and ignores task completion. Since the raters differ by only one point, the procedure does not require a third rater; the two should instead re-read the language-control descriptor together and annotate specific evidence. A score of 3 with the note 'register instability' is the most defensible outcome.
2. What feedback and follow-up task would best address Maya's register problem without discouraging fluency?
Read the worked answer
Feedback should be selective and tied to audience. Rather than marking every Sie/du slip, Keller can replay the role-play and ask Maya to identify the moment the host mother becomes 'Sie' rather than 'du.' Then give two formulaic frames: 'Entschuldigung, dürfte ich ...?' and 'Wäre es okay, wenn ...?' Practice with an audience card: host parent = Sie; host sibling = du. This treats register as a sociolinguistic choice, not a grammar rule to memorize. Keller should also upgrade vocabulary with a short lexical set for meals and health ('zu Abend essen,' 'eine Allergie gegen ... haben'). For fluency, run a 60-second re-performance without correction, then one focused correction round. This avoids turning a level-3 speaking task into a grammar lesson. Follow-up: Maya records herself and marks her Sie/du choices, building metacognition while preserving the strong task completion she already showed.