Curving a High-Stakes Internal Assessment
You are teaching a 60-student postgraduate course, 'Research Methodology', at a Rajasthan university. The first internal test is out of 40 marks. Thirty-eight students scored below 16 (40%). The departmental head advises 'normalization' to avoid poor student feedback, but your learning outcomes require mastery of sampling and hypothesis testing. Item analysis shows 12 items: five had discrimination below 0.15, and two were ambiguous. The syllabus states that internal assessment must reflect criterion-referenced standards. Students are demanding grace marks, and you have one week before the next module. You must decide whether to adjust scores, re-teach, or both, and how to communicate the decision. The head has authority over final internal marks but has left the pedagogical choice to you. You also know that two high scorers missed the test due to illness and will need a re-test.
1. Identify the key decision and the evidence you would use to justify it.
Read the worked answer
The key decision is whether to apply a blanket curve or to treat the test as partially invalid and re-teach/re-assess. The evidence must come from item analysis and construct validity, not from student demands. Five items with discrimination below 0.15 do not separate high and low achievers, so they contribute little to measuring mastery of sampling and hypothesis testing. Two ambiguous items introduce construct-irrelevant variance. A criterion-referenced test should report whether students meet a standard, not rank them. Therefore, the defensible evidence is: (a) the proportion of valid items (5 of 12), (b) the learning outcomes the test was meant to measure, and (c) the syllabus requirement for criterion-referenced assessment. A curve would raise scores but would not fix the measurement problem. The decision should be to re-score after removing or revising the flawed items, then re-teach the weakly measured concepts and offer an equivalent re-test. The two absent high scorers should get the same re-test, not an automatic average.
2. Propose a concrete action plan (including whether to curve) and explain how it aligns with validity and fairness.
Read the worked answer
Plan: (1) Do not curve. (2) Re-score the test out of 36 after dropping the two ambiguous items; for the five low-discrimination items, examine response patterns and either drop them or count them only for feedback, not for marks. (3) Re-teach the concepts those items targeted in a two-hour workshop using worked examples. (4) Offer a re-test to all 60 students plus the two absentees, with parallel blueprints. (5) Use the higher of the two scores, but require students to attend the re-teach session to be eligible. (6) Record the final internal mark against a published rubric. Validity: removing flawed items improves the match between scores and the intended construct. Fairness: all students get the same second chance; no group is penalized for the head's normalization preference. The two absentees are treated identically. This plan avoids the false fairness of a curve, which would reward students who did not master the outcomes and disadvantage those who did.
3. How would you communicate the decision to students and the head?
Read the worked answer
To students: present a short, anonymized item-analysis summary (e.g., 'two questions had ambiguous wording; five did not distinguish prepared from unprepared students'). Explain that the test will be re-scored out of 36 and that a re-test will be offered after a re-teach workshop. State clearly that no grace marks will be given because the syllabus requires criterion-referenced assessment. Offer office hours for individual concerns. To the head: frame the decision in terms of institutional risk and validity. Say that a curve would mask poor item quality and could be challenged under the criterion-referenced syllabus. Provide the item statistics and the re-teach plan. Ask for approval of the re-test schedule and the revised marking scheme. Keep a written record of the meeting. The communication should be transparent, non-defensive, and focused on learning outcomes rather than on student popularity.