Switch course ▾
Session 2. Writing Good Test Items and Rubrics
Objective. Write clear test items that match an outcome, repair flawed items and build a four-level rubric with observable descriptors.
Pupils quickly learn that what is tested is what matters. A poorly written item gives you false information: a pupil may guess right, or know the idea but be defeated by confusing wording. A rubric that says 'good' or 'weak' gives the pupil nothing to aim at. This session shows how to write fewer, better items, and rubrics that two teachers can read in the same way.
Key ideas
Start from the outcome
- An item is good only if it tests the outcome. If the outcome says 'explain', an item that asks pupils to 'name' tests something easier.
- Make a small blueprint: a table with each outcome, the type of item and the marks. A 20-mark test might give 8 marks to the key outcome and 4 to each of the others.
- Test what you taught, in the way you taught it. A surprise format measures exam skill, not learning.
Multiple-choice items
- Write the stem as a clear question that can be answered before looking at the options. Avoid negatives such as 'Which is NOT ...'.
- Make every wrong option plausible, based on a real pupil error, and keep all options the same length and style.
- Avoid 'all of the above', clues in grammar, and a correct answer that is always the longest or always C.
Questions that show thinking
- Short-answer and problem items show reasoning: 'Solve and explain each step' reveals more than the answer alone.
- Write a mark scheme before the test, for example 4 marks: 1 for the set-up, 1 for the method, 1 for the answer, 1 for the unit or explanation.
- Give clear instructions and the marks for each question, so pupils know how much to write.
Rubrics pupils can use
- Choose 3 to 5 criteria and 3 or 4 levels. More than that is hard to use and to read.
- Describe what you can SEE at each level, with numbers where possible: 'gives a reason for 2 of 3 steps', not 'good reasoning'.
- Show the rubric to pupils before the task. Try it on two or three old pieces of work and adjust the wording.
From outcome to a tested item
One item, one idea
A good item tests one idea, so a wrong answer can be interpreted. If a word problem has a long text, a new context and three-step maths, a pupil who fails could have failed in any of the three. For a Grade 7 class where some pupils read Russian less easily, a short plain sentence keeps the item about the maths and not about reading.
Use your blueprint to balance. If the key outcome is worth 40 per cent of the unit, about 40 per cent of the marks should test it. Put easier items first so every pupil gets a start, and keep the layout clean so that crowded pages do not cost marks.
Try the rubric on real work
The first draft of a rubric is a guess. Before you use it, take two or three old pieces of work, one strong and one weak, and mark them. If you hesitate between two levels, the descriptors overlap or are unclear. Rewrite the words, not the mark.
Pupils can help. Show the class two anonymous samples and ask what makes one stronger, then use their words to write the top descriptor. Pupils who helped write 'gives a reason for each step' know what to do. Revise the rubric after every use: note any criterion where most pupils scored the same, because it may not be separating anything.
Examples from the classroom
The word problem that tested reading
Scene. Malika's Grade 7 pupils in Tashkent score badly on a ratio word problem of eight lines about a market stall. When she reads the answers, she sees that many set up the ratio correctly but misread 'per kilogram'. She cuts the text to two sentences, and the next class of 30 gets the same maths right: 22 pupils set it up correctly.
Insight. Cutting extra words made the item test ratio, not reading, so the result could be trusted.
Distractors that talk
Scene. Aziz writes this Grade 5 item: 'Which is larger, 3/8 or 3/5?' A) 3/8, because 8 is bigger. B) 3/5. C) They are equal, because both have 3. D) Cannot tell. In a card poll 14 of 30 pupils choose A and 6 choose C. Each wrong card shows him a different idea, so he plans two short fixes: one with paper strips and one on same-size parts.
Insight. Distractors based on real errors turn a quick poll into a map of pupils' thinking.
A rubric with words you can see
Scene. Dilnoza teaches Grade 9 Russian-medium literature in Samarkand. Her first rubric line says 'Good argument'. After marking two old essays she changes it to: 3 = gives a claim and two reasons, each with a quotation; 2 = a claim with one supported reason; 1 = a claim with no support. Two colleagues then mark the same essay and differ by only one level.
Insight. Descriptors you can count or see make marks more consistent between markers.
Common pitfalls
- Writing items quickly the night before, without checking them against the outcome.
- Using 'all of the above' or an answer that is always longest, which gives clues without testing knowledge.
- Writing rubric levels with words like 'excellent' and 'poor' that only repeat the grade.
- Using a long rubric with eight criteria that no pupil reads.
Fix the Items, Build the Rubric
- In pairs (10 min): read the five flawed items in the Activity Kit. For each, name the flaw and write a better version.
- Compare with another pair (5 min) and agree one rewrite for items A, B and C.
- In the same pairs (15 min): complete the rubric for a Grade 6 water-cycle paragraph with three criteria and three levels. Write descriptors using only observable words.
- Swap rubrics with another pair (10 min). Mark the two sample paragraphs with their rubric and note every word that was unclear.
Your facilitator will guide you through this activity in the session.
Questions to discuss
- Which flaw was the hardest to see in your own style of writing items?
- Which rubric words were unclear when someone else used them?
- What will you change in your next test before you print it?
Reflect
- Which item on my last test would not survive today's checklist?
- Do my pupils know what a good answer looks like before they start?
- Which colleague could try my test before I use it?
Try it tomorrow
- Take one item from your next test and ask a colleague to answer it without options. Note any answer you did not expect.
- Rewrite one 'good/average/poor' rubric line using something you can count or see.
- Show pupils the rubric before the task and ask them to underline the top level.
Further reading and links
Match the item to the outcome, base wrong options on real errors, and describe each rubric level in words someone else can see.