A 1.5-Band Gap Between Two Raters
A response is double-marked. Rater A awards 7.0, Rater B awards 5.5. The script is 291 words and addresses the prompt, but the second body paragraph drifts onto an adjacent issue and, in its final sentence, appears to contradict the position stated in the introduction. Organisation is strong: clear progression, well-managed paragraphs, effective referencing. Vocabulary shows good range with occasional imprecise choices. Grammar shows a wide range of structures with frequent minor errors that do not impede reading. The coordinator must decide what band to record and whether a third marker is needed.
1. Which criterion is driving the 1.5-band disagreement? Explain why the other criteria cannot explain it.
Read the worked answer
Task response.
- Coherence and cohesion is described as strong in the evidence, so neither rater has room to place it at 5.5 - that criterion is effectively agreed at the top of the range.
- Lexical resource and grammatical range are both described as good with non-impeding errors; again, no reasonable reading puts either of those at 5.5.
- The only feature that could plausibly be scored at 5.5 is the drifted, self-contradicting body paragraph. A paragraph that addresses an adjacent issue and then contradicts the stated position is the classic signal that the response 'addresses the task only partially' and that ideas are not consistently developed in support of the position.
So the disagreement is not about the whole script - it is one criterion, one paragraph. Rater A has likely read the script holistically and let strong language and organisation pull task response upward; Rater B has applied the task-response descriptor strictly and let it pull the whole band down.
2. Propose the recorded band and show the arithmetic. Explain why simple averaging of 7.0 and 5.5 is wrong.
Read the worked answer
Best-fit per criterion first, then combine.
- Task response: 5 - partially addressed, one paragraph off-task and a contradictory closing claim.
- Coherence and cohesion: 7 - clear progression, well-managed paragraphs.
- Lexical resource: 7 - good range, occasional imprecision.
- Grammatical range and accuracy: 7 - wide range, frequent minor non-impeding errors.
Mean = (5 + 7 + 7 + 7) / 4 = 6.5, recorded as 6.5.
Why not (7.0 + 5.5)/2 = 6.25: that treats each rater's overall band as an independent measurement of the same quantity. They are not - each rater has produced a profile, and the profiles differ in one place. Averaging whole bands produces a number that matches no profile at all and hides the fact that three criteria are agreed. Averaging also rewards the lenient rater and penalises the strict one equally, when the evidence shows one of them is wrong about one criterion, not half-wrong about everything.
3. What should the coordinator do next, and what would the third marker be asked to judge?
Read the worked answer
Order a third marking, with a narrowed brief.
The brief should not be 'what band is this essay?' - that invites another holistic impression and a third number to average. The third marker should be asked one question: does the second body paragraph develop the stated position, or does it address a different issue and contradict the position? That is the only live dispute.
If the third marker agrees the paragraph is off-task, task response is 5 and the recorded band is 6.5. If the third marker reads the paragraph as a legitimate concession that returns to the position, task response is 7 and the band is 7.
Two further actions: (1) the drifting paragraph and the contradictory closing sentence should be logged as a coaching point - position must be consistent to the last sentence of the last paragraph, not just stated at the start; (2) if this rater pair diverges repeatedly on task response specifically, the fix is descriptor training on partially-addressed responses, not tighter averaging rules.