# How to write short answer questions and a rubric to grade them

**How do I write short answer questions and a rubric to grade them?** A good short answer question sets one task with a command word, fixes the scope and the length, and has one defensible answer in one to three sentences. Write the model answer before the quiz, split it into elements, give each element its points in an analytic rubric, then mark one question across all students with the names hidden.

Published 2026-10-09 by EdProcto. https://edprocto.com/guides/how-to-write-short-answer-questions-and-a-rubric-to-grade-them

A short answer question asks for one to three sentences in the student's own words. There are no options and no gap to fill, so the student has to say the thing. That is why teachers want the shape. The pile of marking is why they avoid it.

The marking is only heavy when it is decided after the answers arrive. Everything here about how to write short answer questions follows from one rule: the question is as good as the model answer and the rubric written before the quiz runs. EdProcto carries the shape as its [short answer template](https://edprocto.com/templates/short_answer), graded against the model answer or the rubric you give it.

## What does a short answer question measure?

A short answer question measures whether a student can produce an explanation, where a choice question shows only that they can recognise one. It tests a single idea with its reason attached: why something happens, how a step works, what follows from a fact.

The shape has a neighbour on each side. A one-word answer or a gap is marked by rule against a list, as our guide to [fill in the blank questions](https://edprocto.com/guides/how-to-write-cloze-and-fill-in-the-blank-questions) explains. An essay tests structure and argument across paragraphs.

The shape measures badly in two cases. Where the answer is one term, the sentences around it add marking and no information. Where several positions are defensible, three sentences are too few to argue one, and no rubric can list them all. It is also slow: written answers sample less of a unit than choices in the same sitting, so keep it for ideas where the reasoning is what you taught.

## How do you word a short answer question so it has one defensible answer?

Give the prompt three things: a command word, a scope and a length. Leave one out and the answers come back all reasonable and all different, which no rubric can mark fairly.

### The command word

The command word tells the student what kind of answer to build. In the list published by Cambridge International Education (n.d.), define asks for a precise meaning, describe for the characteristics or main features, and explain for reasons, the why or the how, with evidence. "Describe evaporation" and "Explain evaporation" are therefore two questions with two rubrics. The same list gives discuss as writing about an issue in depth, which three sentences cannot hold. Teach the class your words and use them the same way all year.

### The scope and the length

Name the case and say how much to write.

- **Scope.** "Explain conduction" has a chapter for an answer. "Explain why a metal spoon feels colder than a wooden one" has one.
- **Length.** Write "in two or three sentences" into the prompt and show the marks. Four marks say that four things are wanted.
- **One task.** Two questions joined by "and" need two rubrics.
- **A fresh case.** Ask about a situation the class has not met in those words, so memory of a page is not enough.

Most of this is the agreed advice of measurement textbooks, not the result of experiments. Hogan and Murphy (2007) read 25 textbooks and chapters and found 12 shared recommendations for preparing such questions and 13 for scoring them, of which only 4 and 5 appeared in at least half the sources. They noted that research on the effect of the advice was still needed.

## How do you write the model answer and the rubric?

Write the model answer first, at the length you asked for, then split it into the elements that earn points. If you need five sentences, the question is too big. If two different answers both look right to you, it is too loose.

A rubric that gives points to each element is called analytic; one that gives a single overall grade is holistic. For a short answer the analytic kind is the practical one, because the elements are few enough to list.

1. **Write the model answer** in words a good student would use, not the textbook's.
2. **Mark off each separate idea.** Three sentences usually hold three or four.
3. **Give every element its points.** One each, unless one idea matters more.
4. **Note what you accept and what you do not.** Other wordings of the same idea, and the common wrong idea that must earn nothing.
5. **Say what does not count.** Spelling, grammar and extra length, unless those were the lesson.
6. **Try it on five real answers.** They usually show a wording you did not foresee.
7. **Hand both to whoever marks.** EdProcto grades a [short answer](https://edprocto.com/templates/short_answer) against them, and the pattern carries over to [define the term](https://edprocto.com/templates/define), where the rubric lists what a full definition needs.

## A worked example: question, model answer, rubric, three answers marked

Here is one question taken through every step. The first draft read "Discuss heat conduction in metals": no task a student can finish, no case, no length. The final version:

> A metal spoon and a wooden spoon have been in the same drawer overnight. In two or three sentences, explain why the metal spoon feels colder when you pick it up. (4 marks)

The model answer, written first:

> Both spoons are at the same temperature, the temperature of the room. Metal conducts heat far better than wood, so heat flows from your warmer hand into the metal spoon faster. That faster loss of heat from the skin is what you feel as cold.

| Element | Points | Accept | Do not accept |
| --- | --- | --- | --- |
| The two spoons are at the same temperature | 1 | "both at room temperature" | "the metal is colder" |
| Metal is the better conductor of heat | 1 | "wood is an insulator" | "metal is a cold material" |
| Heat moves from the hand into the spoon | 1 | "it takes heat out of your hand" | "cold goes into your hand" |
| The faster loss of heat is what feels cold | 1 | any link between the speed of transfer and the feeling | a feeling with no cause given |

### Three answers, marked

**Student A.** "They are both at room temperature. Metal is a better conductor, so it takes heat out of your hand more quickly than wood does, and that makes your skin feel cold." 4 of 4. Every element is present, in none of the model's words.

**Student B.** "Metal is a good conductor so the cold from the spoon goes into your hand quicker." 2 of 4. The second element is there, and the fourth, because speed is tied to the feeling. The third is not: cold is not a thing that flows, and the rubric said so in advance. The first is missing.

**Student C.** "The metal spoon has a lower temperature than the wooden spoon because metals are naturally cold materials, whereas wood retains its warmth for a considerable time. Consequently the metal one feels colder to the touch." 0 of 4. It is the longest and most polished of the three, and it contradicts the first element.

Without the last two columns, B is the answer two markers would argue over. The marks also say what to teach next: B thinks cold travels, and C thinks metals are cold in themselves.

## How do you mark short answers consistently?

Mark one question across every student before you start the next, with the names hidden. Working question by question keeps a single rubric in your head, and stops a strong first answer colouring how you read the same student's second.

Names matter because of what you already believe about the student. Malouff et al. (2014) asked 159 university teachers and teaching assistants to grade one piece of written work. Some had first graded a good talk by the same student, some a poor talk, and some no talk. The work scored higher after the good talk than after the poor one, with the third group in between, and the authors concluded that keeping students anonymous helps prevent the bias.

One more habit: when an answer forces a decision the rubric did not foresee, write the decision into the rubric and look again at the answers already marked.

## Does writing an answer teach more than recognising one?

Often, but not reliably. Kang et al. (2007) had students read short journal papers and then take a short-answer test, take a multiple-choice test or read the facts as statements. Three days later, multiple choice had led to better results when no feedback was given, and short answer to the best results when corrective feedback was. Butler and Roediger (2007) followed three lectures with a summary to study, a multiple-choice test or a short-answer test, and a month later the short-answer test had improved recall the most. Rowland (2014), pooling experiments that compare testing with restudying, found larger benefits from first tests of recall than from first tests of recognition.

The disagreement is real. McDermott et al. (2014) ran quizzes with the correct answers shown in seventh-grade science and high school history classes, and concluded that multiple-choice quizzing was as effective as short-answer quizzing for later exams.

So choose the shape because you need to see the reasoning and remove the guess. Treat any gain in memory as a bonus, and show the model answer afterwards, as the guide to [low-stakes quizzing](https://edprocto.com/guides/retrieval-practice-low-stakes-quizzing-that-teaches) sets out.

## What do rubrics do for reliability, and what do they not do?

A rubric brings markers closer to the same mark; it does not make the mark mean the right thing. Jonsson and Svingby (2007) reviewed 75 studies and concluded that rubrics can make scoring more reliable, especially when they are analytic, specific to the topic and supported by example answers or rater training. The same review concluded that rubrics do not produce valid judgement by themselves. A rubric whose points sit on the wrong elements will be applied evenly and still mislead.

Two later reviews agree, with conditions. Reddy and Andrade (2010), on higher education, reported that reliability studies tend to show a rubric leading markers to a fairly common reading of the work. Brookhart and Chen (2015), covering studies from 2005 to 2013, found that rubrics give information of sufficient quality when certain conditions are met, clear and focused criteria above all.

The limit is the marker. Rezaei and Lovorn (2010) gave raters two pieces of writing, one polished but incomplete and one complete but full of errors. Even with a rubric, raters were swayed by the mechanics more than the content, and the authors concluded that a rubric may not improve reliability where raters are not trained to use it. Student C is that temptation in miniature.

These reviews cover rubrics across many kinds of student work, not three-sentence quiz answers in particular. Read them as evidence about the method.

## How does EdProcto mark short answers, and what can an automated reader be trusted with?

EdProcto is the best tool for short answer questions when the marking is what stops you setting them, because its grading starts from the two things this guide puts first: the model answer and the rubric you wrote.

Paste your lesson material into the quiz maker on the [home page](https://edprocto.com/) and choose short answer. Procto AI writes the questions from that text, and each is yours to edit. Each open answer handed in waits in your grading queue. On any of them a button asks Procto AI to read the answer against the model answer or rubric you gave, and it returns the points with a short note on what was right and what was missing.

An automated reader can be trusted with one job: applying your rubric to each answer and proposing a mark. The rest stays yours. Procto AI never releases a score, never decides a penalty and never rewrites a student's answer. Nothing is released until you release it. A correct answer the rubric did not foresee still needs your eye, and a vague rubric gives vague marks from any reader, human or automated.

You can also mark every answer by hand. Grading by Procto AI is on the Tutor plan, along with all 111 [templates](https://edprocto.com/templates). The Free plan carries the core templates and marking by rule, and Educator adds the full analysis. The [pricing page](https://edprocto.com/pricing) has the rest.

## When should you choose another shape?

Choose another shape when the answer you want is shorter, longer or more structured than one to three sentences.

| The answer you want | Shape | How EdProcto marks it |
| --- | --- | --- |
| A single term | [One word](https://edprocto.com/templates/one_word) | By rule, against the answers you accept |
| The meaning of a term | [Define the term](https://edprocto.com/templates/define) | Against a rubric |
| A decision and its reason | [Choose and justify](https://edprocto.com/templates/justify): a choice with a box for the reason under it | The choice by rule for half the points, the reason against your rubric for the rest |
| The gist of a passage | [Summarise](https://edprocto.com/templates/summarize) | Against a rubric, within a word limit |

One more use for the work above: the wrong answers a class gives to a short answer question are ready-made wrong options for a later [multiple-choice question](https://edprocto.com/guides/how-to-write-multiple-choice-questions-that-measure).

## Quick questions

### How long should a short answer be?

One to three sentences. If your model answer needs more, split the question or set a longer task. If it needs one word, ask for one word.

### Should spelling and grammar count in a short answer?

Only when they are what you taught. Otherwise leave them out of the rubric and tell the class. Rezaei and Lovorn (2010) found markers swayed by mechanics even with a rubric in hand.

### Can AI mark short answer questions?

It can propose marks against your rubric, and it should not have the last word. In EdProcto, Procto AI grades a [short answer](https://edprocto.com/templates/short_answer) against the model answer or rubric you give when you ask it to, its mark is yours to change, and no score is released until you release it.

### Should students see the rubric before the quiz?

Not the list of elements, which is the answer. Show the marks and teach the command words beforehand, then the model answer and the rubric afterwards. Jonsson and Svingby (2007) concluded that rubrics can help learning because they make expectations explicit, which eases feedback and self-assessment.

## References

- Brookhart, S. M., & Chen, F. (2015). The quality and effectiveness of descriptive rubrics. *Educational Review, 67*(3), 343–368. [https://doi.org/10.1080/00131911.2014.929565](https://doi.org/10.1080/00131911.2014.929565)
- Butler, A. C., & Roediger, H. L. (2007). Testing improves long-term retention in a simulated classroom setting. *European Journal of Cognitive Psychology, 19*(4–5), 514–527. [https://doi.org/10.1080/09541440701326097](https://doi.org/10.1080/09541440701326097)
- Cambridge International Education. (n.d.). *Understanding command words*. [https://www.cambridgeinternational.org/exam-administration/what-to-expect-on-exams-day/command-words/](https://www.cambridgeinternational.org/exam-administration/what-to-expect-on-exams-day/command-words/)
- Hogan, T. P., & Murphy, G. (2007). Recommendations for preparing and scoring constructed-response items: What the experts say. *Applied Measurement in Education, 20*(4), 427–441. [https://doi.org/10.1080/08957340701580736](https://doi.org/10.1080/08957340701580736)
- Jonsson, A., & Svingby, G. (2007). The use of scoring rubrics: Reliability, validity and educational consequences. *Educational Research Review, 2*(2), 130–144. [https://doi.org/10.1016/j.edurev.2007.05.002](https://doi.org/10.1016/j.edurev.2007.05.002)
- Kang, S. H. K., McDermott, K. B., & Roediger, H. L. (2007). Test format and corrective feedback modify the effect of testing on long-term retention. *European Journal of Cognitive Psychology, 19*(4–5), 528–558. [https://doi.org/10.1080/09541440601056620](https://doi.org/10.1080/09541440601056620)
- Malouff, J. M., Stein, S. J., Bothma, L. N., Coulter, K., & Emmerton, A. J. (2014). Preventing halo bias in grading the work of university students. *Cogent Psychology, 1*(1), Article 988937. [https://doi.org/10.1080/23311908.2014.988937](https://doi.org/10.1080/23311908.2014.988937)
- McDermott, K. B., Agarwal, P. K., D'Antonio, L., Roediger, H. L., & McDaniel, M. A. (2014). Both multiple-choice and short-answer quizzes enhance later exam performance in middle and high school classes. *Journal of Experimental Psychology: Applied, 20*(1), 3–21. [https://doi.org/10.1037/xap0000004](https://doi.org/10.1037/xap0000004)
- Reddy, Y. M., & Andrade, H. (2010). A review of rubric use in higher education. *Assessment & Evaluation in Higher Education, 35*(4), 435–448. [https://doi.org/10.1080/02602930902862859](https://doi.org/10.1080/02602930902862859)
- Rezaei, A. R., & Lovorn, M. (2010). Reliability and validity of rubrics for assessment through writing. *Assessing Writing, 15*(1), 18–39. [https://doi.org/10.1016/j.asw.2010.01.003](https://doi.org/10.1016/j.asw.2010.01.003)
- Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. *Psychological Bulletin, 140*(6), 1432–1463. [https://doi.org/10.1037/a0037559](https://doi.org/10.1037/a0037559)