Double marking a source-based paper: discrepancy beyond tolerance
At a European Baccalaureate school, the History written paper is marked out of 60: Section A source analysis (20) and Section B essay (40). Candidate R's paper is independently marked twice. Marker A awards 13 + 25 = 38; Marker B awards 15 + 29 = 44. The department's tolerance is 5 points, so the 6-point gap triggers a third marker. The school's written protocol says the final mark is the median of the three totals; language errors may reduce the Section B communication mark by no more than 2, but must not affect Section A. Marker A applied a 2-point language deduction; Marker B applied 1. Marker B also praises four own-knowledge facts about unemployment insurance that do not appear in the sources; Marker A says these facts crowded out evaluation of provenance. The third marker, marking blind, gives 14 + 27 = 41. The grade boundary is 42. The candidate's teacher asks whether the paper can be averaged to 41.5, rounded to 42.
1. Apply the protocol. What is the final mark, and may the school average or round the paper up to 42? Justify your decision.
Read the worked answer
The protocol says median, not mean. Sorting the totals gives 38, 41, 44; the median is 41. The school cannot average 41 and 44 to 42.5, nor average all three to 41, then round up. Rounding rules apply to the recorded mark, not as a device to cross a boundary; on the facts there is no half-mark to round. The final mark is therefore 41/60, below the grade boundary of 42. The third marker's blind status and the arithmetic should be recorded. If the school's written policy allows rounding only of exact .5 totals, it is not engaged. A moderation meeting may check that the third marker applied the same mark scheme, but it should not negotiate a mark to reach a grade. The teacher's request should be refused, with the breakdown explained.
2. Resolve the two criterion disputes: own-knowledge facts in Section A and the language deduction. What should the moderation meeting decide?
Read the worked answer
Own knowledge: in Section A, the construct is source analysis, not a memory test. Facts absent from the sources can be credited only when they help test provenance, purpose, reliability or usefulness. If the four unemployment-insurance facts were used to challenge the sources, they are relevant; if they replaced evaluation, Marker B's 15/20 is too generous. The third marker's 14/20 is consistent with that distinction. In Section B, relevant own knowledge can earn AO1, but it cannot repair a weak argument. Language: the policy permits up to 2 marks off Section B communication, not content, and never Section A. Marker A's -2 is within the cap; Marker B's -1 is more lenient but not automatically wrong. The meeting should compare other scripts to check consistency, not split the difference. If language is assessed against the candidate's L2 level, minor errors should not be treated as content errors. Neither dispute changes the median.
3. What feedback and appeal information should the candidate receive?
Read the worked answer
Feedback should give the three totals and the median rule, then criterion-level comments: source evaluation, use of own knowledge, essay argument and language. It should say that the mark was not averaged or rounded because the protocol does not permit it. The candidate should be told what would have raised the mark: evaluating provenance and purpose in Section A, and using outside facts only to test the sources. On appeal, the school should check whether the third marker was blind, whether the language cap was applied consistently, and whether the median was correctly calculated. A disagreement about professional judgement is not by itself a procedural error. If the appeal finds that Marker B's Section A mark was inflated, the median may remain 41; if the arithmetic is wrong, correct it. Do not promise a grade change before the review.