Comparability Across Two Classes Teaching the Same Task
Two teachers deliver the same School Assessment task: a 5-minute persuasive speech in the target language on an environmental issue, marked out of 20 against three criteria (content, language, delivery). Teacher X's class average is 15.8; Teacher Y's is 12.4. Both classes are mixed-ability and were allocated without streaming. Teacher Y argues that their class 'genuinely performed worse' and that the difference should be accepted. Teacher X suspects they mark more generously. A sample of six scripts from each class, cross-marked by both teachers, shows Teacher X's marks are on average 2.1 higher than Teacher Y's on the same scripts, with the largest gaps on the 'delivery' criterion.
1. Interpret the cross-marking data and advise on what should happen before results are finalised.
Read the worked answer
The cross-marking data is the key evidence. On the same six scripts, Teacher X marks 2.1 higher on average. That is a systematic difference in standards, not a difference in student ability - the scripts are identical, so the only variable is the marker.
The largest gaps on delivery suggest that criterion is the least clearly defined. Delivery criteria are often vague ('confident', 'engaging'), which invites idiosyncratic interpretation. Content and language criteria tend to be anchored in observable features.
Teacher Y's claim that 'the class performed worse' is not supported by the cross-marking: it may be true, but the data cannot separate class ability from marker severity. The 3.4 average gap is too large to attribute to allocation alone without evidence.
Before finalisation:
- Reconcile the two markers on the sample, using the performance standards, until they agree on what a 12, 15 and 18 looks like.
- Clarify the delivery criterion with observable descriptors (e.g., pace, pronunciation accuracy, use of pauses, eye contact) to reduce interpretation drift.
- Re-mark or adjust using the reconciled standard, and document the process for moderation.
The principle: comparability of standards across classes is a school responsibility, and cross-marking data must be acted on, not explained away.
2. The school proposes to apply a statistical adjustment raising Teacher Y's class marks by 2.1 to match Teacher X. Evaluate this.
Read the worked answer
A blanket statistical adjustment is not appropriate and should be rejected.
The cross-marking showed Teacher X is 2.1 higher than Teacher Y on the same scripts. That means the likely error is in Teacher X's marks being too high, not Teacher Y's being too low. Raising Teacher Y's class by 2.1 would move marks in the wrong direction and widen the discrepancy.
Even if the direction were correct, a uniform statistical adjustment is problematic because:
- It is not evidence-based for individual students. A student at 11/20 and a student at 18/20 may not be affected equally by marker severity; the error may be concentrated at particular mark ranges.
- It obscures the real problem, which is criterion ambiguity, especially in delivery.
- It is hard to defend at moderation without a documented re-mark against clarified standards.
The better process: reconcile standards, re-mark the affected scripts against the agreed standard, and only then compare class distributions. If after re-marking a genuine class difference remains, that is a finding to report, not to correct statistically. Adjustments should follow evidence about individual performances, not cohort averages.