Adding Ten Marks: An End-of-Term Assessment Problem

Decision 1 - adding 10 marks: rejected, but for a reason worth stating precisely.

Adding a constant to every score does not make the assessment fairer; it makes it uninterpretable. It preserves the rank order exactly, so it tells nobody anything new about who needs help - the 12 learners below 30% stay at the bottom. What it does destroy is the link between the score and the construct: a parent, the headteacher or the next teacher can no longer read '22/50' as evidence about what this learner can do in Social Studies. A blanket adjustment also makes his class incomparable with the other JHS 3 class, which matters if the school uses SBA for placement or reporting.

If the paper genuinely was unfair - items from outside the taught syllabus, ambiguous wording, miskeyed answers, questions testing topics not covered - the defensible remedy is an item-level one: remove or re-score the flawed items for everyone, recalculate out of the reduced total, and record why. That is defensible because it is transparent, it is based on evidence, and it can be explained to a parent.

Decision 2 - entering the class average for seven learners: not permissible.

Estimating scores and then recording them as real data is falsification of an assessment record. It misleads the learners, their families, the headteacher and, if SBA is used externally, it misleads the system; it also removes any reason for those seven learners to be assessed properly. A missing record is an administrative failure, and the honest responses are: report the gap, use other evidence that already exists (exercise books, project work, end-of-term exam performance), and where nothing exists, record the learner as not assessed and let the school's weighting policy handle it - with the headteacher's written knowledge.

Underlying principle. School-Based Assessment is evidence of learning, not a form to be completed. Once a teacher starts manufacturing data to hit a deadline, every other judgement he makes becomes suspect, and the learners who most need support are exactly the ones whose scores get invented.

He must not interpret the mean until he has described the test. A mean of 38% is a fact about the interaction between this test and these learners; it says nothing on its own about ability or about teaching quality.

Evidence to assemble before the meeting:

1. A content-coverage map. For each of the 50 items, the sub-strand or topic it tests and whether that topic was taught, when, and for how long. Any item on untaught content is not a measure of his learners; it is a measure of the booklet.
2. Item facility (p-value). Proportion correct per item. Items below about 0.20 may be miskeyed, badly worded, or simply not taught; items above 0.90 carry little information.
3. Discrimination. Whether strong learners got the item right more often than weak learners (compare upper and lower thirds on total score). Negative discrimination is a red flag for a miskey or an ambiguous stem.
4. Distractor analysis. Are wrong answers spread across options, or does nearly everyone pick the same distractor? A heavily chosen distractor often reveals a specific misconception worth re-teaching - or a confusing option.
5. Language and reading load. Social Studies items with long stems or unfamiliar vocabulary may be testing reading, not the concept, especially for learners who are still developing academic English.
6. Triangulation. The term's class tests, exercise books, attendance and the previous term's results. If the class tests showed the same weakness, the 38% is a real learning gap. If the class tests were fine, suspect the paper.

Then interpret. A mean of 38% with a top score of 74% and a floor of 12% is a very wide spread. A class where the strongest learner gets three-quarters and the weakest gets an eighth is not a class with one problem; it is a class with at least two - a group with genuine gaps in foundational content and a small group who are coping. That argues for differentiated re-teaching, not a whole-class response and certainly not a flat mark addition.

What to say to the department. Present the coverage map and item data first, then the conclusion: some of the low mean is a test-construction problem and some is a learning problem, and both need fixing.

Weeks 1–2: diagnose and re-teach the top three weaknesses.

• From the item analysis, select three sub-strands where facility was low but the content was taught. Re-teach each in a single 20-minute cycle: a concrete example, a worked model, guided practice, then three items the learners do alone.
• Pre-test each sub-strand with five items at the start of the week and the same five at the end, so learners see their own movement.
• Give back the exam scripts for 15 minutes in class. The rule is that learners do not just look at the score - they record, on a small grid, which questions they lost marks on and why (did not know / misread / ran out of time / guessed). The grid converts a score into a plan.

Weeks 3–4: differentiated small-group work.

• Group learners by the sub-strands they missed, not by total score: the 12 lowest are not one group.
• Use peer tutors from the top band for mechanical recall (dates, definitions, terms); Mr. Mensah takes the conceptual work with the middle band, who are the group most likely to move.
• Weekly 10-item quiz, marked and returned within 48 hours with a written comment on one thing to do differently.

Weeks 5–6: examination technique under time.

• One full timed section per week, with explicit teaching of elimination, reading the stem before the options, and how long to spend per item.
• Analyse the weakest three items again; if they are still failing, the teaching, not the learners, needs changing.

Feedback design that matters. Comments should be about the task and the strategy ('you chose the option that was true but did not answer the question'), not about the learner ('you are weak'). Public posting of scores, ranking or comparison between learners should stop - it drives the bottom group away from the subject exactly when they need to stay in it.

Evidence for the HoD at week 6: pre/post scores per sub-strand, the number of learners moving out of the lowest band, and a short note on which items were withdrawn from the original paper and why.