Bimodal Unit Test and Three Suspect Items
The eighth grade U.S. history team gave a 30-item multiple-choice and two constructed-response unit test to 112 students across four sections. The overall mean was 71 percent. Section 2 averaged 62 percent; section 4 averaged 79 percent; the middle sections clustered near 72. The item analysis flagged three items. Two used long, multi-clause stems ending in 'EXCEPT' and had discrimination indices near zero. On the third, 61 percent of students chose D, but the key is B. Nine students with IEPs for reading scored below 50 percent, and their constructed responses were mostly blank or one sentence. The team meets Monday to decide whether to reteach, curve, or move on. Report cards close in ten days, and the next unit builds directly on this content.
1. What do the item and section data actually tell you, and what do they not tell you?
Read the worked answer
Start by separating what is solid from what is confounded.
Solid: an overall mean of 71 percent is not a crisis by itself. But nine students with reading IEPs scoring below 50 percent, with mostly blank or one-sentence constructed responses, is a signal rather than noise - it points at reading and writing load on the assessment itself, not only at content knowledge.
Confounded: the three flagged items. A miskeyed item (61 percent chose D, key is B) means some students who understood the content were marked wrong, so the section means are partly artifacts. The two 'EXCEPT' items with near-zero discrimination are likely measuring reading stamina and attention to negation rather than the history standard. Section 2's 62 percent versus section 4's 79 percent could be a teacher effect, a schedule effect (section 2 may be first period), or a scoring effect - you cannot tell from a mean alone.
What the data cannot tell you is whether students cannot explain the content or cannot write about it. Before Monday, pull three blank constructed responses and interview those students orally.
2. Recommend a course of action for Monday's meeting, including what to do about grades.
Read the worked answer
Recommend in this order.
- Rescore before you interpret. Remove or credit the three flawed items for all 112 students and recompute. Announce why: the items were flawed, not the students. This protects trust in the assessment system, which matters more than the points.
- No curve. A curve would bury the very signals - the reading-IEP gap and the writing gap - that the team needs to act on.
- Two days of targeted reteach on the two or three standards where the corrected data show real weakness, not a reteach of the whole unit. Nine days is not enough for more, and because the next unit builds on this content, embed the reteach in the opening lessons of the new unit rather than pausing for a separate week.
- A replacement opportunity on those standards for any student below 70 percent after reteach - same standards, different items - capped at the original weight so grades reflect learning rather than one Tuesday.
Grades should stand for what was measured correctly. The remedy for a bad measure is a better measure, not inflation.
3. How would you adapt the next unit for the nine students with reading IEPs without lowering the content bar?
Read the worked answer
Keep the standard and change the access route.
Before reading: pre-teach eight to ten terms with visuals and a quick matching check, not a glossary handed out at the door. During reading: provide a version at accessible readability that carries the same core content, plus a one-page summary with the same facts; these students are not excluded from the primary source, they get a supported version alongside it. During writing: sentence starters for each part of the constructed response, and the option to plan orally before drafting.
Build in a low-stakes checkpoint midway through the unit - three items in the same format as the test, including one 'EXCEPT' item - so students practice the format that tripped them up and the teacher gets early evidence while there is still time to act.
The key judgment: an oral response option is an access accommodation, not a lower bar. The student must still make the same historical argument; only the modality changes. The IEP team should confirm which accommodations are already documented before anyone promises them.