The elegant essay that answers a different question
The L2 written examination includes a writing paper marked with a grid that separates task achievement (10) from language (10). The task asked candidates to argue for or against making school uniforms compulsory, referring to two of the three documents supplied. One script of about 700 words is error-free, wide-ranging and well organised, but after two opening sentences about uniforms it argues that 'all rules limit the freedom of the individual' and never returns to the documents. The final sentence reads: 'This is why uniforms should not be imposed.' The first corrector awards 16/20, citing range, accuracy and fluency. The second awards 9/20, citing failure to address the task. The permitted tolerance is 3 points, so a third correction is required. The candidate is at a grade boundary: seven points on this paper moves the overall subject mark across a threshold that affects university admission.
1. What exactly do the two correctors disagree about, and how does the grid resolve it?
Read the worked answer
They agree on the observable facts - the script is accurate, fluent and largely irrelevant - and disagree about what the paper is measuring. That is a weighting dispute, not a marking error, and the grid already answers it.
Apply each half separately.
Task achievement (out of 10). The criteria here concern relevance to the prompt, coverage of the required sources, development of a position, organisation and register. The script has a clear thesis, coherent paragraphs and an appropriate register, so it is not a zero: it does address the question nominally and states a position. But it ignores the two-document requirement entirely and substitutes a general liberty argument for the specific debate. Descriptors of that kind typically place it in the lower third: 3/10, possibly 4/10 if the band rewards organisation strongly, and no higher.
Language (out of 10). Range, accuracy, cohesion. On the evidence described this is high: 8/10, perhaps 9/10.
Total: 11–12/20. The second corrector is closer to the truth; the first cannot reach 16 without crediting task achievement at 7 or 8, which the evidence does not support. The first corrector has in effect marked the language twice.
A useful diagnostic question for the feedback report: would the corrector have given a linguistically weaker but equally off-topic script 16? If not, the mark is a fluency halo.
2. What should the third corrector be given, and how should the final mark be arrived at?
Read the worked answer
What to give them. The script, the task, the source documents and the grid. Also the two correctors' criterion-level comments, because that is the audit trail and it shows where the divergence originates. What they should not be given is the two totals, and certainly not the information that the candidate is at a boundary. Anchoring on 16 and 9 would turn the third correction into a compromise between two numbers rather than an independent judgement.
How to mark. Independently, criterion by criterion, writing a band and a reason for each half of the grid before converting to a total. On the analysis above, 11 or 12 is the expected landing point.
How to finalise. The aggregation follows the procedure in force - commonly the third marker's mark, or the mean of the two closest marks. The coordinator should state which rule is being applied before the totals are compared, so that the rule is not chosen after seeing the result.
What matters beyond this script. If the two comment sheets show that one corrector never addressed task achievement at all, the problem is standardisation and it will recur across the cohort. The coordinator should sample ten scripts for the same pattern and, if it is found, arrange a criterion-level re-marking of the affected paper before results are released - not after an appeal.
3. Should the effect on the candidate's university admission influence any part of the decision?
Read the worked answer
No, and the answer has to be operational rather than rhetorical. Two things must be kept apart: the marking of the script and the communication of the result.
On the marking, the boundary is not a criterion. Double correction with a defined tolerance exists precisely so that a mark is not the product of one marker's temperament or of a desired outcome. Marking a script up to protect an admission, or down to protect a quota, corrupts the only evidence the certificate carries. The candidate is entitled to the mark the grid supports, not to the best mark the grid could be argued to support.
On consistency, the legitimate version of the fairness worry is this: whatever band is applied to this script must be applied to every script with the same profile in the same session. If the school's practice has been to reward linguistic flourish regardless of task, correcting it for one candidate at a boundary is unjust. So the coordinator should check other scripts with a similar divergence and apply the same standard - which may mean correcting some marks upwards as well as downwards.
On communication, the candidate should receive the criterion-level feedback, the information that a third correction took place, and the appeal route. That is the proper response to the consequence. It is also worth recording that the seven-point swing arose from a divergence well beyond tolerance, which is a quality-assurance finding about the marking, not about the candidate.