A Flawed Test and a Grade Deadline
Mr. Santos teaches Grade 8 Science to 48 learners. He constructed a 40-item quarterly test and built a table of specifications: 24 items on Remembering, 10 on Understanding, and 6 on Applying. Item analysis produced these results: Item 12, p = 0.92, D = -0.15; Item 20, p = 0.30, D = 0.35; Item 33, p = 0.15, D = 0.05. The class mean is 19 out of 40, and the test counts for 30% of the quarterly grade. Grades are due in five school days. Mr. Santos proposes adding 10 points to every learner's raw score so the mean becomes 29. A parent has already emailed asking why her daughter, who usually earns 85, got 21/40. The department head asks him to justify both the test and the proposed adjustment before the grades are finalized.
1. Interpret each item's difficulty and discrimination, and decide whether to retain, revise, or drop each item. Justify each decision.
Read the worked answer
Item 12 (p = 0.92, D = -0.15) is easy but negatively discriminating, which is a warning sign, not merely an easy item. Higher-achieving learners were more likely to answer it incorrectly. The first step is to inspect the key and the distractors: a miskey or ambiguous wording is more likely than a genuinely defective concept. If the key is wrong, rescore all learners on that item and consider excluding it from the total.
Item 20 (p = 0.30, D = 0.35) is difficult but positively discriminating. It separates learners who understood from those who did not, so retain it unless it is off-specification or unreadable. A high-quality hard item is useful.
Item 33 (p = 0.15, D = 0.05) is very hard and does not discriminate, which suggests it is poorly written, off the table of specifications, or tests something that was not taught. Drop it from scoring or revise and pilot it.
Decide the rule before computing grades and apply it to all learners, because dropping items changes the denominator.
2. Evaluate the proposal to add 10 points to every learner's score. What would you do instead to produce defensible grades within the deadline?
Read the worked answer
Adding 10 points to everyone is not defensible. It inflates all scores equally, compresses the differences between learners, and leaves the flawed instrument in place. It also creates a misleading record: a parent comparing 21/40 with 31/40 is not being told what the learner actually knows. The proposal responds to the mean, not to the evidence.
A better sequence within five days: (1) review the key and wording of the flagging items, especially Item 12; (2) recompute scores against the set of valid items, with the denominator adjusted and a written rationale; (3) use the other components, such as performance tasks and quizzes, to corroborate the picture; (4) offer one targeted reassessment of the same competencies if policy allows. Communicate transparently with parents: what was found, what was changed, and how the final grade was derived. A criterion-referenced adjustment can be defensible if it is documented, applied uniformly, and tied to specific evidence, but a blanket bonus is not.
3. What does the table of specifications imbalance imply for the validity of this test, and what is one concrete change for next quarter?
Read the worked answer
With 24 of 40 items at Remembering, the test overrepresents recall and underrepresents Understanding and Applying. This is construct underrepresentation: a high score may show that a learner memorized facts, not that the learner can explain or use science ideas. The low mean does not by itself prove the learners failed the intended outcomes, because the instrument was misaligned with them.
A concrete change is a two-way table of specifications that distributes items across cognitive levels and content areas, for example roughly 40% Remembering, 40% Understanding, and 20% Applying, with item formats that require explanation, data interpretation, and transfer to a new situation. Then pilot the items, compute difficulty and discrimination, and revise before the test counts. This also makes the parent conversation easier: the grade can be tied to stated competencies and item evidence, rather than to a single flawed score.