Skip to content

Test the outcome, not the trivia. Start from what students should be able to do and write each question to show whether they can. In tertiary education that usually means applying, analysing and evaluating, not just recalling. A question that asks students to pick the most likely diagnosis from a short case tests far more than one that asks them to name a drug class, and it’s much harder to answer by memorising notes.

Use scenarios to test judgement. A good vignette gives students facts to interpret, not the answer in disguise. Keep it to three to five short sentences, include only the details that matter to the decision (perhaps one realistic distractor detail), and finish with a single, focused question such as “What is the most appropriate next step?” or “Which principle best applies?” This is how professional licensing exams in medicine, nursing and law are written, and it’s equally useful in teaching, business and IT.

Make the stem a complete question. A well-written stem passes the “cover test”: a student who covers the options should be able to answer it. Put everything shared by the options into the stem, ask a direct question ending in a question mark, and cut anything that doesn’t help students answer it. Avoid fill-in-the-blank stems and stems ending in a colon.

Never word the question negatively. “Which of the following is NOT…” and “All of the following EXCEPT…” fail the cover test, reward test-taking skill over knowledge, and are easy to misread under exam pressure. Ask for the best or most appropriate answer instead. A negative inside the scenario itself (“she has not taken her medication for two days”) is fine; it’s the question that must be positive.

Write distractors from real misconceptions. The best wrong answers are the ones a student with a specific gap in understanding would actually choose. That makes the question discriminate between students who understand and those who don’t, and the distractor feedback tells you exactly which misunderstanding each wrong answer reveals. Implausible “filler” options just reduce a five-option question to a two-option guess.

Keep options alike and give away nothing. Options should be the same kind of thing (all diagnoses, all doses, all actions), similar in length and grammatically consistent with the stem. Avoid giveaways: a correct answer noticeably longer or more qualified than the rest, a word repeated from the stem, or absolutes such as “always” and “never” that students learn to rule out. Make sure exactly one option is defensibly correct and that options don’t overlap. Avoid “all of the above” and “none of the above”, which test exam technique rather than knowledge. List numeric options in order.

Fewer good options beat more weak ones. Three or four plausible options make a better question than five with a couple of obvious fillers. If you can’t think of a plausible fifth distractor, choose fewer options for the set.

Check every question before students see it. Confirm the keyed answer is correct and current for your discipline and region, that the scenario is realistic, and that no question gives away another question’s answer. AI-generated questions, however carefully checked against writing rules, are a strong first draft, not a finished exam. For high-stakes assessment, have a colleague answer the set blind, and after the exam check each question’s difficulty and how well it discriminates (most LMSs report this) so you can retire or revise the weak ones.

Use MCQs for what they do well. They’re efficient, reliable and quick to mark, and they suit frequent low-stakes quizzes, revision and self-testing as well as exams. They don’t show whether a student can explain, argue or perform a skill, so pair them with written, practical or case-based tasks when the outcome calls for it.