A multiple-choice question is only measuring what it claims to measure if every wrong answer is genuinely plausible to someone who doesn't know the material. Most item-writing flaws come from violating that one condition, and they're almost always accidental — the item writer knows the answer, so implausible distractors don't register as implausible to them.
Implausible distractors turn a knowledge test into a guessing test
If three of four options are obviously wrong to anyone with basic familiarity with the topic, the item is effectively a two-option question wearing a four-option costume, and test-takers who don't know the material can often eliminate their way to the right answer anyway. Every distractor should represent a plausible misconception someone with partial knowledge might actually hold — not a random unrelated fact.
Grammatical cueing gives away the answer
If the stem ends in "an" and only one option starts with a vowel, you've cued the answer through grammar rather than content. The same happens when the correct answer is noticeably longer or more precisely worded than the distractors, since item writers tend to unconsciously hedge the correct answer with more careful language. Read every option against the stem in isolation and check that all of them are grammatically and stylistically interchangeable.
"All of the above" and "none of the above" are worth avoiding entirely
"All of the above" can often be answered correctly by recognizing that just two of the options are true, without needing to evaluate the third at all — partial knowledge produces a correct guess. "None of the above" has the opposite problem: it tests whether a test-taker can identify that everything else is wrong, which is a different and often harder skill than recognizing what's right. Both options are well-documented in assessment literature as sources of construct-irrelevant variance — noise that has nothing to do with the knowledge being tested.
The stem should ask a complete question on its own
If a test-taker has to read all four options before understanding what's actually being asked, the stem is incomplete. A well-formed stem poses a clear, complete question or statement, and the options simply complete or answer it — this alone eliminates a large share of construct-irrelevant reading-comprehension load from an exam that's supposed to be testing subject knowledge.
The practical check before publishing any item
- Could someone eliminate at least two distractors through general test-taking savvy alone, without subject knowledge?
- Does the stem read as a complete question without needing the options to make sense of it?
- Is every distractor a genuine, specific misconception, not a random unrelated fact?
- Would a subject-matter expert agree the correct answer is unambiguously correct, with zero defensible argument for a distractor?
None of this requires new content expertise — it requires a second, mechanical pass over items that are already written, looking specifically for these patterns rather than re-reading for accuracy alone.