Writing Good Exam Questions: A Practical Guide Using Bloom's Taxonomy

Question writing is the invisible craft of teaching. A well-written question reveals precisely what a student does and does not understand; a poorly written one rewards guessing, punishes reading speed, or tests something you never meant to test. This guide covers the practical rules that separate the two.

Use Bloom's taxonomy as a difficulty dial

Bloom's taxonomy classifies cognitive demand into levels — remember, understand, apply, analyse, evaluate, create. Its practical value for question writers is as a dial: the same content can be asked at any level, and the level, more than the content, sets the difficulty.

• Remember: 'State the formula for kinetic energy.'
• Understand: 'Explain why doubling speed more than doubles kinetic energy.'
• Apply: 'A 1,200 kg car travels at 25 m/s. Calculate its kinetic energy.'
• Analyse: 'Two cars have equal kinetic energy but different masses. Which is travelling faster, and how do you know?'
• Evaluate: 'A student claims a heavier vehicle is always more dangerous in a collision. Assess this claim using energy principles.'

A balanced paper deliberately distributes questions across levels — typically weighted toward apply and analyse for exam preparation, since that is where most exam marks live. If every question on your worksheet is a 'state' or 'define', students can score well while understanding little.

The classic flaws that leak marks

Double-barrelled questions

'Explain how enzymes work and why temperature affects them' is two questions wearing one mark allocation. Students answer half, markers argue about partial credit. Split it, and allocate marks to each part explicitly.

Cue leakage

Questions often accidentally contain their own answers — grammatical agreement in multiple-choice options, a diagram label that gives away the mechanism, or an earlier question whose answer feeds a later one. Proofread the paper as a cheating student would: what can I infer without knowing the content?

Implausible distractors

In multiple-choice questions, the wrong options do the teaching work. Distractors should each represent a real, known misconception — the answer a student gets by forgetting to square the speed, or by confusing mass with weight. Options that are obviously absurd reduce a four-option question to a coin flip. Your students' past wrong answers are the best distractor bank you will ever have.

Reading load masquerading as difficulty

A long, convoluted stem does not make a question more rigorous; it makes it a reading test. Keep stems lean: context in one or two sentences, then a clean command word. Save complexity for the thinking, not the parsing.

Command words are a contract

'State', 'describe', 'explain', 'compare', and 'evaluate' each promise a different depth of answer, and students are explicitly trained to respond to them. Using 'explain' when you will accept a description — or marking a 'describe' answer down for not explaining — breaks that contract and teaches students that command words are noise. Decide the expected answer first, then choose the command word that matches it.

Write the mark scheme at the same time

The fastest way to discover a flawed question is to write its mark scheme. If you cannot list what earns each mark, students cannot know either. A useful discipline: for every question, note the level you intended (apply, analyse, and so on) and check the mark scheme actually rewards that level rather than rewarding recall of a model answer. Over a term, this habit does more for the quality of your assessments than any question bank — though it also makes every question you bank considerably more reusable.