The fluent candidate who talks the most
Four S6 candidates sit an eight-minute group discussion as part of your school-based oral assessment. The topic: 'S3 students should spend one afternoon a week on community service instead of a second PE lesson.' You and a colleague score independently.
A: 9 turns, 41% of the talk time; interrupts B once; twice answers a question nobody asked; links explicitly to another candidate twice. B: 4 turns, 12% of the talk time; hesitates ('the - the thing is'); introduces 2019 flood-relief volunteer figures; opens one turn with 'Going back to what A said about timetabling …'. C: 6 turns, 18%; reads from a prepared script; repeats the university-application point three times. D: 7 turns, 29%; paraphrases B's point accurately; asks C a direct question; says 'there is many benefits'; misplaces the stress in 'voluntary' but is understood without effort.
Your colleague wants A ranked first because of fluency and range.
1. Argue against ranking A first, using the specific evidence recorded.
Read the worked answer
Talk time is a measurement, not an achievement. A's 41% is the largest share, but the criteria the panel uses include interaction and communication strategies, and that is where A's record is weakest: one interruption, two turns that answer questions nobody asked, and only two explicit links to other speakers.
Read the two irrelevant turns carefully. They are monologues - probably well-formed ones, which is why they sound impressive. In a discussion, a turn that ignores what was just said is a failure to respond, however fluent the grammar. The interruption is the same failure in a louder form: it takes a turn away from someone else.
Fluency and range are real strengths and should be credited under delivery and language. The error would be letting them stand in for the whole profile. A candidate who speaks 41% of the time is also shaping the group's opportunity structure: B's 12% is partly a consequence of A's dominance, which means the panel cannot read B's low output as low ability without discounting for it.
Suggested order at this stage: D first (broad evidence across criteria), B second (high-quality contributions, explicit building), A third, C fourth. That order should be provisional until the panel has listened to A's two monologues again to check whether they contain any response to the preceding speaker.
2. B and D are close. What actually distinguishes them, and how do you decide?
Read the worked answer
What B shows: content of high quality - the 2019 figures are evidence, not opinion - and an explicit link that shows real listening. What B does not show is range across functions: four turns is thin evidence, and the hesitation, while normal, means we cannot judge fluency confidently.
What D shows: accurate paraphrase (the clearest single indicator of listening in the whole transcript), an inclusive move in asking C a direct question, and a fair share of talk time. D gives the panel evidence on more criteria.
How to decide: ask which candidate's score would change most if the group had been different. Give B a fair hearing by imagining B in a group without A: on this evidence B would probably speak more and the hesitation would matter less. That argues for crediting the quality of B's four turns generously rather than treating the low count as a ceiling. But we cannot award marks for contributions we did not hear.
Outcome: D ranks above B because D's evidence is broader and comparable in quality; B sits clearly above A and C. Record the reasoning in the moderation note - 'B's low talk time partly attributable to A's dominance' - so an appeal or a re-mark has something concrete to review. Do not resolve the closeness by averaging the two markers' ranks; resolve it by returning to the criteria.
3. How should the panel treat 'there is many benefits' and the misplaced stress in 'voluntary'?
Read the worked answer
Two different categories of error, and neither is decisive on its own.
'There is many benefits' is a subject-verb agreement error. One occurrence in seven turns is a slip. If it recurred three or four times, it would be a pattern and would count under language patterns. The panel should therefore count occurrences before deciding, not react to the first one heard.
Misplaced stress in 'voluntary' is a pronunciation issue, and the relevant test is intelligibility: the notes record that D was understood without effort, so this is a minor item. It would become significant if a listener had to reconstruct the word, or if the misplacement recurred across the candidate's lexical range.
The principle: assess the effect on the listener and the consistency of the error, not the number of errors. A single striking error is more memorable than four unremarkable ones and therefore more likely to distort a holistic impression - which is precisely why the panel uses recorded evidence and criteria rather than memory.
Practical step: before scoring, each marker independently notes errors and their frequency, then compares. If the two markers disagree on whether D's grammar is a 'pattern', that disagreement is about evidence, and it can be settled by replaying the recording rather than by splitting the difference in the final mark.