Every Milestone reading unit is checked the way a test publisher checks an exam. Below is what we check, and three real flaws our checks caught in our own units, with what we changed.
What every unit is checked for
Every answer key recomputed
A script re-derives each printed answer from the unit's source data. If a key and a question disagree, the check fails and the unit goes back for a fix.
Tested for answer patterns
Keys are tested for letter runs, repeating cycles, “longest answer” tells, card-sort positions and a second form that copies the first.
Questions that need the passage
Questions are tried without the passage. If a Grade 3–5 student could answer it from general knowledge or a wording tell, it is rewritten.
Facts traced to the source page
Every factual claim in a passage is checked against the opened source page, and the supporting quote is kept on file.
Worked examples: flaws we caught and fixed
Described without naming the unit. Each fix was confirmed in the unit's source files before we wrote it up here.
Example 1: Answers a student could find by position
What we found. In one unit's card sort, 10 of 12 cards landed in the right column if a student simply glued each card straight across from where it was printed. The same check found post-checks whose answer letters repeated the pre-check's order.
Why it matters. A student who spots the pattern scores well without reading, and a pre-check to post-check gain stops meaning growth.
What we changed. We re-deal the printed order: the words and the correct answers stay the same, only where they sit changes, and the answer keys move with them because they are generated from the same data. After the re-deal, the straight-across shortcut scores 6 of 12 — no better than guessing. The position check is now part of the gate every new or revised unit must pass before release.
Example 2: A question answerable without the passage
What we found. A character-traits question asked which trait a character showed. Two options were negative traits and one was a feeling, so the answer was the only positive trait. In another unit, the word “probably” appeared in all seven wrong “no clue” options and in no correct answer.
Why it matters. Both are rules a student can learn once and apply forever. Neither is the reading skill the unit teaches.
What we changed. We replaced the giveaway options with positive traits the story does not show, so only the story can decide, and rebalanced “probably” so it no longer marks a wrong answer. Each changed item was re-validated.
Example 3: An invented detail in a task card
What we found. A summarizing task card said two characters thought a classmate's idea was “great.” The story only says they agreed to try it.
Why it matters. A card that adds a feeling the text never states quietly teaches students to invent evidence — the opposite of the skill.
What we changed. The card now says exactly what the story says, the answer key's reason was rewritten to match, and a new check in that unit fails any card that credits a character with words or feelings the story does not contain.
What these checks do not promise
These are classroom practice resources, not standardized or norm-referenced tests, and they carry no ministry approval. Checks reduce errors; they do not make a resource perfect. If you find a mistake, email us and we will fix it.