How to Write Good Multiple-Choice Questions (with Bloom's Taxonomy)

To write good multiple-choice questions, define the knowledge or reasoning you need, write one defensible key and build plausible alternatives. Then check whether wording, option length or missing conditions allow a learner to answer for the wrong reason.

This guide walks through how to write good multiple-choice questions that hold up in a real classroom, from the wording of a single stem to spanning the levels of Bloom’s taxonomy.

Start with the stem

The stem is the part of the question that poses the problem. A weak stem forces the Student to read all the options before they understand what is being asked. A strong stem can almost be answered before the options are even shown.

A few principles keep stems sharp:

• Pose a complete question, not a fragment. “Which of the following best describes photosynthesis?” works better than “Photosynthesis is…” because the second version turns each option into a separate sentence-completion puzzle.
• Put the bulk of the wording in the stem. If every option repeats the same phrase, move that phrase up into the stem. Shorter options are quicker to scan and reduce reading load.
• State it positively where you can. Negatives (“Which is NOT a mammal?”) are sometimes necessary, but when you use them, make the negation impossible to miss by emphasising it.
• Keep it free of clutter. Background detail that does not affect the answer just slows the reader down and rewards stamina rather than knowledge.

Here is the difference in practice:

• Weak: “Water…” — a) boils at 100 °C at sea level, b) freezes at 10 °C, c) is denser as a gas, d) contains three hydrogen atoms. Each option is a separate sentence to parse, and they do not even test the same property.
• Stronger: “At approximately standard atmospheric pressure, at what temperature does pure water boil?” — a) 80 °C, b) 90 °C, c) 100 °C, d) 110 °C. One clear question, four homogeneous options, answerable before the choices even appear.

If a colleague can read your stem and tell you what the correct answer should be without seeing the options, you are on the right track.

Apply that test to one question from your next lesson. Draft the stem first, ask a colleague for an answer before revealing the choices, then check whether every wrong option represents a mistake a learner might actually make. If you want to test the finished item with a class, create a free teacher account and share a short Quiz by link or QR code; the response pattern will show which distractors did useful work.

Write distractors that do real work

Distractors affect whether the item distinguishes useful response patterns. A good distractor is not a throwaway. It represents a plausible misconception, a common error in reasoning, or a half-remembered fact. When a Student picks it, you have a possible explanation to check with their reasoning.

To write distractors that earn their place:

• Base them on real mistakes. Think about the errors your Students actually make, the step they skip, the rule they over-apply, the definition they confuse. Those make the most diagnostic options.
• Keep them homogeneous. All options should belong to the same category and be roughly the same length and grammatical form. A single option that is longer, more detailed, or phrased differently quietly signals that it is the answer.
• Make every option plausible. Joke options and obviously absurd choices waste a slot and inflate the score without measuring anything.
• Aim for three or four options. Beyond that, you spend effort inventing filler distractors that no one seriously considers. Three solid options beat five weak ones.

That final number is a practical rule, not a universal psychometric law. Published guidance for constructing multiple-choice questions stresses one best answer, plausible distractors and review, while a broader review of best practices in examination-item construction explains why distractors should distinguish knowledge from test-taking skill. High-stakes examinations need local validation and item analysis beyond the drafting advice in this article.

Avoid unintended cues

Even a well-written question can leak the answer through small tells. These cues let a test-wise Student score points without knowing the material, which undermines the whole exercise.

Watch out for:

• Grammatical giveaways. If the stem ends in “an”, an option starting with a consonant sound is ruled out by grammar alone. Keep the grammar consistent across every option.
• Length tells. The longest, most qualified option is too often the correct one, because we instinctively over-explain the truth. Trim the right answer or pad the distractors so they match.
• Absolutes and hedges. Options containing “always” or “never” tend to be false; options with “usually” or “may” tend to be true. Use these words evenly or not at all.
• Answer-position habits. If your key is “C” more often than chance, learners notice. Vary the position of the correct option.
• Overlapping options. When two choices can’t both be wrong, attentive Students deduce the answer by elimination rather than knowledge.

Span the levels of Bloom’s taxonomy

Use Bloom’s taxonomy to describe the intended cognitive task: recall, interpretation, application or analysis. The label does not establish difficulty or quality, and adding a scenario does not automatically make an item analytical.

• Remember. Straight recall of facts, terms, or definitions. “What is the chemical symbol for sodium?” Useful for building a foundation, but easy to over-rely on.
• Understand. Asking a Student to interpret or explain in their own terms. “Which statement best paraphrases this passage?”
• Apply. Putting a rule to work in a new situation. Give a short scenario and ask which principle applies, or have the Student calculate a result.
• Analyse. Breaking something apart: spotting the flaw in an argument, comparing two cases, identifying which piece of evidence supports a claim.

To climb the ladder, anchor the question in a stimulus, a passage, a data table, an image, a worked example, and ask the Student to interpret or apply it. Such a task can also involve retrieval; the processes are not exclusive. A well-built bank that mixes these levels tells you not just whether Students know the facts, but whether they can reason with them. This is the same principle behind formative assessment that genuinely improves learning: the format is only as good as the thinking it provokes.

Worked MCQ revision: ask about the interval, not an impression

Suppose the learning goal is comparing average temperature decrease per minute over two intervals. Give a fictional cooling record rather than relying on what a learner thinks cooling should look like. These figures illustrate item design; they are not a claimed physical law or a laboratory result.

A vague draft asks: “When is it cooling fastest?” It leaves unclear whether the learner should identify an instant, an interval or the coldest reading. Revise it to: “During which five-minute interval is the average temperature decrease per minute greatest?” Use three comparable options: 0–5 minutes, 5–10 minutes, or the decreases per minute are equal in both intervals.

The key is 0–5 minutes: (80 − 50) ÷ 5 = 6 °C per minute, compared with (50 − 35) ÷ 5 = 3 °C per minute. The question defines a positive decrease; it does not ask learners to choose the numerically greatest signed rate of change. That wording avoids confusing −6 and −3 with the intended magnitude.

A wrong choice is still only a hypothesis about reasoning. Ask the learner for a step, especially before assigning an intervention. The table makes two plausible mistakes visible without claiming everyone selecting an option used that method.

Pilot with a changed key before reusing the item

For a parallel item, use 90 °C at zero minutes, 70 °C at five and 30 °C at ten. The average decreases are 20 ÷ 5 = 4 and 40 ÷ 5 = 8 °C per minute; the key becomes 5–10 minutes. This changes the correct interval while preserving the calculation demand, so success cannot rely solely on remembering the first option position.

Keep the two tables and keys in an external item record. If you need Skolina’s advanced per-question statistics for the pilot, check Starter, Pro or eligible institution access on the plans page. An option distribution helps you decide what to inspect, but it cannot establish which calculation each learner performed.

Choose the right question type

Multiple choice is the default, but it is not always the best fit. Matching the format to the learning goal often does more for question quality than wordsmithing a stem. A platform like Skolina offers eight question types, and each suits a different kind of task:

• Single choice for one clearly correct answer among distractors.
• Multiple choice (several correct) when several options are valid. Check the actual scoring rule; selecting some correct options does not necessarily earn partial credit.
• True/false for quick concept checks, though it is coarse and easy to guess.
• Matching for relationships, terms to definitions, causes to effects.
• Fill in the blank to test recall in context without offering cues.
• Numeric for calculations where an exact value is the answer.
• Ordering for sequences, processes, timelines, and steps.
• Open response for genuine production, where you accept that grading takes more effort.

A short Quiz can mix formats when each collects evidence needed for the learning goal; variety alone does not guarantee engagement or understanding. If you are setting one up from scratch, our walkthrough on building an online quiz for your classroom covers the mechanics. And when you have a course or a PDF to start from, you can draft a first set of questions with AI and refine from there. A word of honesty on that last route: the AI generation runs on OpenAI’s models, so that text is processed outside the EU, and a draft is never a finished assessment. Human review of every stem and distractor stays essential.

Common mistakes to avoid

Even experienced teachers fall into a handful of recurring traps. A quick checklist before you publish:

• “All of the above” and “none of the above.” These reward partial knowledge or test-taking tactics more than understanding. One known false option rules out “all of the above”; recognising several true options can also supply a clue without checking every option. One true option alone does not establish that all are true.
• Negative stems left unmarked. A buried “not” or “except” trips up Students who know the material perfectly well.
• Trick questions. Hinging an item on an obscure exception or a deliberately misleading phrase measures attention to traps, not learning.
• Testing trivia. If the fact would not matter outside the test, it probably should not be on it.
• One question, two ideas. If a stem bundles two concepts, a wrong answer tells you nothing about which one the Student missed. Split it.
• Copying the textbook verbatim. Lifting a sentence word for word rewards memorising surface phrasing rather than meaning.
• Skipping the review. Reading your own questions cold, a day later, catches more ambiguity than any rule. Better still, ask a colleague.

A chosen distractor is a hypothesis about a misconception, not a diagnosis. Ask a learner to explain the choice, check whether wording or guessing explains it, then use a parallel question to see whether the reasoning persists. The guide to distractors that diagnose misconceptions shows that sequence in detail.

Putting it into practice

Review the complete item before sharing it. Start with a stem a colleague could answer unaided, build distractors from the mistakes your Students really make, strip out the cues that leak the answer, and deliberately push some items up Bloom’s ladder so you are measuring reasoning and not just recall. Check the key and response format in the actual student view, then read a sample of explanations after the pilot.

You can compare the current Skolina plans and Quiz limits before moving a question bank online. A shared document or paper pilot is often the fastest way to get a colleague’s review first.

The craft takes practice, but it pays off in assessments that actually tell you what your class understands. When you are ready to turn these principles into a working question bank, you can create a free teacher account and start drafting straight away.