Two assessors, a memorised talk and an accommodation
It is oral assessment day: 28 candidates, two assessors (one Nama-speaking, one Haiǁom-speaking), four criteria at 5 marks each. Candidate P delivers a polished three-minute prepared talk on 'My village', fluent and well organised, but when asked two unprepared follow-up questions repeats sentences from the talk and cannot respond when a question is rephrased. Assessor A gives 16/20 and assessor B gives 10/20. Candidate Q speaks with a repaired cleft palate; articulation is unclear, but the follow-up questions show she understood and can repair misunderstandings when asked to repeat. Candidate R uses Haiǁom vocabulary and click realisations, and assessor B wants to deduct for 'non-standard' language. The recordings exist and the head of department must reconcile the marks before results go to the examinations directorate.
1. Resolve the disagreement over P without simply splitting the difference.
Read the worked answer
The prepared talk tests delivery and organisation; the follow-up questions test comprehension and interaction, which are separate criteria. P's memorised script can score well on the first, but the follow-ups show the language is not available for unplanned use: repeating the script is not an answer.
So mark criterion by criterion on the recording rather than averaging two impressions. If delivery carries, say, 8 of the 20, P can keep 6–7 there, while the interaction criteria should be low. The 16 and the 10 are both global impressions, not criterion marks, and neither can be defended against the descriptors.
Replay the recording with both assessors present, marking one criterion at a time. Record the reconciled mark and the reason, and tell the next cohort that the talk may be prepared in advance but the questions will not be.
2. How should Q be assessed, and what must be written down?
Read the worked answer
Assess communicative competence, which is what the criteria describe, not the palate. Q understood the follow-up questions and repaired misunderstanding when asked to repeat - that is exactly the strategic competence the interaction criteria reward. Mark content, organisation and interaction on what she did, and do not let an articulation impression leak into every criterion.
Two safeguards. Make the reasonable accommodation explicit and written: extra response time, permission to ask for repetition, a quiet room, face-to-face seating. Check it against the national policy before the marks are finalised. Then record the accommodation on the mark sheet so that the external moderator sees why the evidence looks as it does.
If intelligibility genuinely blocks communication on a given item, that item scores low - but the cause must be communication failure, not the presence of a disability.
3. Rule on assessor B's objection to R's variety, then set out how the overall moderation should run.
Read the worked answer
On the variety: the criteria assess communication in Khoekhoegowab, and Haiǁom is a variety of it. Regional phonetic realisations and locally current vocabulary are not errors unless the syllabus prescribes one standard for the oral. Deducting for 'non-standard' without a written descriptor is unfair, so B must point to the clause that justifies the deduction; if no such clause exists, restore the marks.
On the moderation: with recordings available, the two assessors should re-mark independently, one criterion at a time, then compare. Where they still differ by more than 2 marks on a criterion, a third assessor decides. Record the reconciled mark and the reason for it.
Do not average away a disagreement that moderation exists to settle: a split with no evidence behind it hides exactly the problem the head of department is meant to fix.