The Assessment That Looked Fine

Warranted inferences:

• The gap is between procedures and explanation, not between "fractions" and "not fractions." Students can compute; they struggle to justify.
• The assessment under-samples the standard's central demand: 30% reasoning items against a 40% blueprint, and zero visual-model items when the standard explicitly names visual fraction models. The 34% figure therefore may overstate or understate the gap - it certainly cannot stand in for the standard.
• The EL gap is roughly constant across clusters: 92−61 = 31 points, 77−44 = 33 points, 39−12 = 27 points. A gap that does not widen as the task becomes more conceptual points toward broad language access rather than a reasoning-specific deficit.

Not warranted:

• "The class can't do fractions" - 89% on procedures contradicts it.
• "The ELs have a conceptual deficit in equivalence" - the flat gap is consistent with vocabulary and item wording, and the test never gave them a visual representation to reason from.
• "Re-teaching the same item types will raise the score" - no error analysis has been done yet, so the cause of the 34% is still unknown.

The plan treats a symptom.

• More problems of the same type re-exposes students to the format that they already failed, without supplying the missing representation or the missing language. If the barrier is "I don't know how to say why 2/3 = 4/6," a seventh worksheet does not help.
• It was chosen from item p-values, not from student work. The team has not looked at a single response. The same 34% may contain three different errors: no explanation attempted, explanation that restates the question, explanation that uses "bigger numbers" instead of a model.
• It ignores the test's own misalignment. Re-teaching to a test that omits visual models reinforces the gap between instruction and the standard.

What to do instead:

1. Error analysis first (one planning period). Pull 12–15 responses across the score range, sort into error categories, and count them. Decide the reteach from the categories.
2. Teach to the representation the standard names. Fraction strips and area models, side-by-side fraction comparisons, and "convince a skeptic" prompts.
3. Supply the language. Sentence frames ("___ and ___ are equivalent because the same amount of the whole is shaded when…"), oral rehearsal in pairs before writing, and a posted word bank (equivalent, numerator, partition, whole).
4. Align the next assessment. Rewrite 6 of the 20 items to reach the 40% reasoning blueprint, with at least two items requiring a visual model.

Run a short, structured diagnostic on a sample of 6–8 EL students, one at a time, and vary one thing at a time.

1. Oral re-interview with the model present. Give the student a fraction strip or area model and ask the same comparison question orally. If the student can explain correctly aloud but not in writing, the barrier is production, not concept.
2. Strip the context. Re-administer two items as bare numbers ("Is 2/3 equal to 4/6? Show how you know") with no word-problem frame. Improvement here points to reading/vocabulary load in the stem.
3. Add a scaffold and see what moves. Re-administer the original items with a glossary of the three or four key terms and a sentence frame. If scores rise sharply, the gap is access; if they do not, the gap is conceptual and needs reteaching.
4. Check the student's own explanation. Ask, "What is the question asking you to do?" in the student's preferred language or plain English. Students who cannot restate the task are failing the reading, not the math.
5. Compare with a matched non-EL sample on the same scaffolded versions. If both groups improve, the items are simply hard; if only the EL group improves, the demand is linguistic.

Use the result to choose between language supports and content reteaching - and do not assume the answer before the diagnostic.