Moderating a Split Oral Mark in Rukwangali First Language
At a secondary school in Rundu, two Rukwangali First Language teachers assess the same Grade 12 learner's individual oral. The task is a six-minute talk on 'Uukwashike wokulima moKavango' (the value of farming in Kavango), followed by four minutes of questions. Assessor A awards 18/25 and notes 'confident, strong examples, engages the audience'. Assessor B awards 12/25 and notes '13 long pauses, 22 fillers, uneven pronunciation of prenasalised consonants, no signposting'. The recording shows that the learner reads about 40% of the talk from cue cards, answers two questions with single words, then recovers in the last 90 seconds with a detailed example of mahangu storage. The department's analytic grid weights content 30%, fluency 30%, pronunciation 20% and interaction 20%. The HOD must reconcile the marks before moderation. No third assessor is available, and the recording is the only evidence.
1. Diagnose why the two assessors diverged and set out a defensible moderation procedure using the recording and the analytic grid.
Read the worked answer
The divergence is not a disagreement about the learner's ability in the abstract; it is a disagreement about which construct each assessor foregrounded. A's notes use global, impressionistic language ('confident', 'engages') and reward content and presence. B's notes count surface features and apply a fluency/pronunciation lens. Because the grid is analytic, neither global impression nor a single deficit can settle the mark.
Procedure:
- Listen once without scoring to establish the overall profile.
- Listen again with the grid, timestamping evidence for each criterion (e.g. 0:00-2:40 cue-card reading; 3:10 one-word answer; 5:20 recovered example).
- Score each criterion independently before combining.
- Compare only evidence, not adjectives.
- Record the final mark and a short rationale for external moderation.
Do not average 18 and 12 as a compromise: that would hide the criteria conflict. The HOD should also check whether A ignored the 40% reading and whether B double-counted pauses under both fluency and pronunciation.
2. Using the grid, propose a mark or narrow range and justify it with timestamped evidence from the recording.
Read the worked answer
Using the grid: content 30% = 7.5 marks; fluency 30% = 7.5; pronunciation 20% = 5; interaction 20% = 5.
Evidence and marks:
- Content 6.0/7.5: the mahangu storage example is specific and relevant, and the talk addresses the title, but the Q&A does not extend the content.
- Fluency 3.5/7.5: 40% cue-card reading and long pauses damage flow, though the final 90 seconds is sustained.
- Pronunciation 3.5/5: intelligible, with uneven prenasalised consonants, but meaning is not impeded.
- Interaction 2.5/5: two one-word answers and limited turn-taking; the learner does not ask for clarification or build on prompts.
Total 15.5/25, so a defensible range is 15-16. The mark should not be 12: that would let fluency and interaction wipe out intact content. It should not be 18: that would ignore the reliance on cue cards and the weak Q&A. The rationale must name the evidence and the weighting.
3. Design a two-week intervention targeting the most damaging weakness, with two measurable success criteria.
Read the worked answer
The most damaging weakness is interaction, because it depresses two criteria (interaction directly and fluency indirectly) and is least likely to improve by accident. Two-week plan:
Week 1: three four-minute pair tasks in which the learner must answer 'why' and 'how' questions with one claim + one example + one consequence. Record and count one-word answers. Week 2: two six-minute talks with cue cards reduced to five keywords, plus a three-minute unseen Q&A with a peer asking follow-ups.
Success criteria:
- No more than two one-word answers in a four-minute Q&A.
- Cue-card reliance under 15% of talk time, measured by a timer.
Pronunciation work is embedded only where it blocks meaning, not as a separate drill. The intervention targets the behaviour that costs most across the grid, not the most visible symptom.