A Mean of 62 and What It Actually Measures
Mrs. Bell's Grade 8 U.S. History unit test produced a class mean of 62%. Item analysis shows: five multiple-choice items with three implausible distractors had 91% correct; four short-answer items requiring a full paragraph had 60% of students scoring 0–1 out of 4, most writing a single sentence or restating the prompt; two items covered a video shown during a fire-drill day when six students were absent. Mrs. Bell has two class periods before the district checkpoint, which covers the same standards. Her instinct is to reteach the whole unit, but there is not enough time, and several students who scored well are already asking for enrichment.
1. What does this test actually tell Mrs. Bell about student learning, and what does it fail to tell her?
Read the worked answer
The test is not measuring content mastery cleanly; it is measuring several things at once:
- The five 91% items carry almost no information. When nearly everyone answers correctly, the item does not discriminate and it should not drive instruction. It may mean mastery, or it may mean the distractors were too easy.
- The short-answer collapse is ambiguous between writing facility and content knowledge. Students writing one sentence or restating the prompt may not know the content, or may not know how to build a paragraph-length explanation. Those two causes require completely different reteaching.
- The two video items produce an invalid comparison. Six students were absent for the source, so their scores on those items reflect attendance, not learning.
What it fails to tell her: which specific concepts are weak. A mean of 62 is an average of items that measure very different things, and acting on the average rather than the pattern is how teachers end up reteaching what students already know while the real gap survives.
2. Given two periods before the checkpoint, what should she reteach, what should she not reteach, and how should she resolve the ambiguity in the short answers?
Read the worked answer
Do not reteach the whole unit. Most of the block would be spent on material the 91% items and the strong performers already demonstrate.
Resolve the ambiguity first, cheaply. In the first five minutes of the next period, re-ask two short-answer prompts orally or in bullet form to a sample of students who scored 0–1. If they can state the content in bullets, the problem is response format or writing stamina, and reteaching content will fix nothing. If they cannot, the problem is conceptual.
Then allocate:
- If it is a writing/format problem: model a paragraph with a claim, two pieces of evidence, and a link; give a frame and have students rewrite one answer. This is a short, high-yield intervention.
- If it is a content problem: reteach only the specific concepts behind the missed short answers, using a sorting task or timeline reconstruction that makes the relationships visible.
- The two video items: do not judge the six absent students on them. Give them a brief alternate exposure (short reading or clip at the start of class) and check separately.
- The students asking for enrichment should be working on a transfer task (e.g., evaluating which cause was most significant and defending it), not sitting through reteach. That is also the group best positioned to help - but as explainers of a specific task, not as tutors for the whole unit.
3. Mrs. Bell will reuse this unit next year. What two changes to the assessment itself does this analysis justify?
Read the worked answer
1. Fix the item writing. Replace the five non-discriminating items with distractors built from plausible misconceptions - the wrong answers students actually gave on this test are the best source. And commit to a fixed format for the short-answer items: if a paragraph is the target skill, say so in the standard and teach it; if content explanation is the target, allow bullet points or a labeled diagram so the score reflects the standard being assessed rather than writing fluency.
2. Remove the attendance confound. Any content delivered in a single, non-recoverable event (a video on a fire-drill day) should either have a text backup students can access independently or should not be assessed on the common test. A test score that partly measures who was in the room is not a valid measure and, with two such items, it is enough to distort a class mean.
A third, optional change worth considering: keep the item analysis from this year and use it next year as a pre-assessment blueprint, so the unit opens with the pattern that took two weeks to discover this time.