The Word Bank Question: Accommodation or Construct Change?
Mr. Okafor teaches five sections of seventh-grade science (148 students total). On the last ecosystems unit test, his third-period class averaged 61% while his other four sections averaged 74%. Third period contains six English learners at WIDA levels 2-3, three students whose IEPs specify read-aloud and extended time, and two students with 504 plans for ADHD. At the data meeting, a colleague proposes giving the entire class a word bank of key terms, converting every four-option item to three options, and reading the test aloud to everyone, so that third period's scores will be 'comparable' to the other sections before the district benchmark window closes. The principal asks Mr. Okafor to justify in writing whatever he adopts, because benchmark scores feed the school's improvement rating and are disaggregated by teacher.
1. Evaluate each of the three proposed supports. Which can be applied to all students without changing what the test measures, and which would make third-period scores non-comparable to the other sections? Justify each decision.
Read the worked answer
The decision turns on naming the construct first. If the unit's target is content understanding (energy flow, food webs), then reading load is incidental; if target vocabulary is itself assessed, supplying definitions removes the very thing being measured.
- Word bank: defensible only for non-assessed academic vocabulary ('compare,' 'describe'). A bank covering the tested terms converts a recall item into a matching item and inflates scores. Keep two banks or label clearly.
- Three options instead of four: raises guessing from 25% to 33%, lowers item difficulty, and changes the score scale. Any cross-section comparison becomes invalid. Reject for benchmark purposes.
- Read-aloud: for the three IEP students and, if documented, the 504 students, it is an accommodation already owed. As a universal support it is arguably construct-neutral for science but must be reported as an accommodation so scores are not treated as identical conditions.
The defensible move is to standardize the reporting conditions across sections and keep the construct fixed, not to engineer equal averages.
2. A parent of a student with no IEP or 504 plan emails requesting read-aloud and extended time for her daughter on the same benchmark. What do you do, and how do you document it?
Read the worked answer
Treat the request as information, not as an entitlement or a refusal. First ask what the parent has observed and whether the difficulty is reading fluency, attention, or test anxiety, and check the student's screening data. If the need appears real and persistent, the appropriate route is a referral for problem-solving team review and, if warranted, a 504 evaluation.
For the immediate benchmark, I would allow read-aloud as an instructional support if it does not affect the reported construct, and I would tag the score so it is never pooled with standard-condition scores. I would not grant it silently: the documentation should record the request, the date, the decision, the rationale, and where the score is stored.
The risk to avoid is creating an informal accommodation system that bypasses evaluation and leaves the student without legal protection later, which is exactly what happens when a teacher quietly reads to whoever asks.
3. What analysis would make third-period's benchmark data actually defensible to the principal and to the district?
Read the worked answer
Raw section averages compare different students, not different teaching. A defensible analysis would:
- Disaggregate by EL status, IEP/504, and prior-year proficiency, so the 61% can be read against the six ELs and five accommodated students rather than attributed to instruction.
- Use matched or growth comparisons - fall-to-winter change for the same students - instead of a single post-test average.
- Run item-level analysis to see whether the gap is concentrated in reading-heavy constructed responses (a language-load problem) or spread evenly (a content problem).
- Report conditions of administration next to every score so nobody mistakes accommodated and standard scores for the same measure.
Finally, I would ask that these scores not be used in teacher evaluation without the same disaggregation. A 13-point gap between sections can be entirely composition, and treating it as teacher effect is the most common misuse of benchmark data.