Item Writing and Review Best Practices
This domain is about producing items that measure the intended construct without introducing irrelevant difficulty. The central principle is construct-irrelevant variance: anything that makes an item harder for reasons unrelated to the knowledge being tested (ambiguous wording, cultural references, unnecessarily complex vocabulary) is a defect. The companion principle is construct underrepresentation: an item that is too easy or too narrow fails to sample the intended skill.
Item-writing rules to name: write a self-contained stem that poses a clear problem; put most of the content in the stem and keep options short; make all options homogeneous in length, grammar, and category; avoid absolute terms ("always," "never") unless the content is truly absolute; avoid negative stems and double negatives, or if unavoidable, capitalize the negative; avoid "all of the above" and "none of the above" because they alter guessing behavior; avoid clang associations (a distractor that echoes a word from the stem); and avoid convergence cues (the longest or most detailed option being correct).
Review process concepts: items pass through content review (accuracy and alignment to the blueprint), editorial review (grammar, clarity, fairness), and sensitivity/fairness review (bias, stereotypes, accessibility). Subject-matter expert (SME) review validates accuracy; psychometric review validates statistical behavior after piloting.
Traps: (1) An item can be factually correct yet still defective for ambiguity. (2) "Improving" an item by adding detail to the key - this creates a length cue. (3) Assuming review is one step; it's a sequence of distinct reviews. (4) Confusing fairness review (bias) with editorial review (clarity).
Rule of thumb: if you can answer the item without reading the options, the stem is well-formed. Worked example: "Which is NOT a benefit of X?" - rewrite as "Which is a limitation of X?" to remove the negative stem.