# Summary questions: testing comprehension in fifty words

**How do I set and mark a summary question?** A summary question gives a passage and a word limit, and is marked fairly when the rubric comes first. List the main ideas of the passage before the quiz, give one point for each in any wording, add a point for using no copied sentence, give nothing for an opinion or a detail, and stop reading at the limit.

Published 2026-10-10 by EdProcto. https://edprocto.com/guides/summary-questions-testing-comprehension-in-fifty-words

A summary question hands the student a passage and asks for its gist in a set number of words: "Summarise the passage in fifty words." Nothing can be guessed: the student decides what matters and says it briefly.

That makes summary questions quick to set and slow to mark, unless the main ideas are listed in advance. EdProcto has the shape as its [summarise template](https://edprocto.com/templates/summarize): a passage condensed into a set number of words, graded against your rubric.

## What does summarising a passage ask of a reader?

Summarising asks a reader to do four things to a text: leave out, generalise, select and construct. In the model of Kintsch and van Dijk (1978), understanding a text already means condensing its full meaning into a gist, and a summary puts that step on paper. Brown and Day (1983) turned the model's rules of deletion, generalisation and construction into four a marker can look for:

- **Deletion.** Drop what is trivial, and what is important but said twice.
- **Superordination.** Put one term in place of a list: pets for cats, dogs, goldfish, gerbils and parrots.
- **Selection.** Use the topic sentence, if the author wrote one.
- **Invention.** If there is none, write your own.

### Which of the rules are hard?

Deletion is easy and invention is hard. Brown and Day (1983) asked 67 students, from fifth grade to college, to summarise two geography texts of about 500 words. Every age group used the deletion rules about nine times in ten or better. Superordination and selection grew with age. Invention was rare among fifth graders, tenth graders used it on a third of the occasions that called for it, and college students on half.

The weaker summarisers worked by what the authors call copy-delete: take each sentence in turn, keep it or drop it, and copy what is kept nearly word for word.

## Does summarising help learning?

As a way to revise, summarising helps less than its popularity suggests, because it depends on how well the student can already summarise. The review of ten learning techniques by Dunlosky et al. (2013) rated summarisation low in utility: it has helped some students on some kinds of test, under limited conditions. Dunlosky (2013) gave teachers the reason: students who cannot yet write a good summary need extensive training first. In the study he describes, pupils had five coaching sessions of about fifty minutes each, and then recalled more of a chapter's important points than pupils with none.

Spirgel and Delaney (2016) ran five experiments and found no evidence that writing a summary beat restudying the text. Students remembered what they put in their summaries better than what they left out, but were poor at picking out the important points.

So summarising is worth teaching and testing, and a weak way to revise for a student never taught it. For revision, see our guide to [low-stakes quizzing](https://edprocto.com/guides/retrieval-practice-low-stakes-quizzing-that-teaches).

## Is a summary a fair test of comprehension?

A summary tests comprehension and writing together, so the mark speaks about reading only if the rubric makes it.

**It sees what choice questions miss.** Head et al. (1989) gave seventh graders a social studies text and found that multiple-choice and summary measures overlapped little in the kinds of comprehension they assessed.

**Writing leaks into the mark.** Brown and Day (1983) had to tell their raters to ignore style, because the older students also wrote better.

**Length and copying mislead.** In Yu (2008) the longer summaries were rated the poorer ones, and in Keck (2006) second-language writers leaned on near copies.

The rubric therefore pays for ideas found, withholds a point for lifted sentences, and stops at a limit.

## How do you choose the passage and set the word limit?

Choose a passage from your own lesson with four to six main ideas, and set the limit just above the length of your own summary of it. That is our working rule: we found no experiment that fixes a ratio.

For a quiz question we suggest:

- **120 to 200 words,** written or adapted by you from what was taught.
- **One of each trap:** a list to generalise, an example to delete, and a main idea that no single sentence states.

### What a word limit forces

A limit forces the student to generalise. Brown and Day (1983) also asked for a 60-word summary of each text, the length experts had needed. Every age group kept to it, and from seventh grade up the students replaced lists with a general term more efficiently under the limit than without one.

Our working view of the two errors: too tight, and the ideas cannot all be said, so the mark records which one a student happened to drop; too loose, and copy-delete passes. In the example below, a limit of 110 words would let a student lift six sentences whole and touch every idea. Write your model summary first, then round its length up. The [summarise template's page](https://edprocto.com/templates/summarize) suggests a third of the passage as a fair target.

In EdProcto the editor has a box labelled "Word limit (0 for none)". The student sees a running count under the answer box, and "over the limit" beside it once the limit is passed. Nothing stops the typing and the count plays no part in the mark, so put the rule on length in your question and your rubric.

## What goes in a summary writing rubric?

A summary writing rubric lists the main ideas, one point each, and little else.

1. **Write the model summary yourself,** inside the limit. If you cannot, the limit is wrong.
2. **List its main ideas,** usually four to six, at one point each, with the wordings you accept and what earns nothing.
3. **Add one point for own words:** no sentence lifted whole from the passage.
4. **Give nothing for an opinion, an example or a detail,** and take nothing off. They cost the words they fill.
5. **Rule on the limit.** Ours: words past it are not read.
6. **Leave spelling and grammar out,** unless they were the lesson.
7. **Try it on five real summaries.** In EdProcto the [summarise](https://edprocto.com/templates/summarize) editor then takes the list in its "Rubric (what earns the points)" box, beside one for the model answer.

## A worked example: one passage, a 50-word limit, four summaries marked

The passage, 148 words, written for a geography lesson:

> A city is usually warmer than the countryside around it, and the gap is widest in the hours after sunset. On a calm, clear night the difference can reach several degrees.
>
> Asphalt, brick, concrete and dark roofing soak up sunlight all day and release it slowly through the evening. Anyone who has crossed a car park barefoot on a summer night has felt it. Fields and woods behave differently. Plants draw water from the soil and release it through their leaves, and that evaporation cools the air much as sweat cools the skin. A street with few trees gets little of this cooling. Cars, buses, air conditioners and factories add heat of their own.
>
> None of this is fixed. Street trees, flat roofs covered with soil and grass, and roofs painted white all keep a district cooler in summer. A city, in other words, can be made cooler.

Our own summary of it ran to 46 words, so the limit is 50, a third of the passage.

| Main idea, 1 point each | Earns nothing |
| --- | --- |
| 1. A city is warmer than the country around it | "cities are hot" |
| 2. Building materials store the day's heat and release it later | "asphalt is hot" |
| 3. A city has few plants, so little cooling by evaporation | how plants cool, with no link to the city |
| 4. Engines and machines add heat | |
| 5. More plants and paler roofs cool a city | "councils should plant trees" |

A sixth point is for own words. Words after the fiftieth are not read. The degrees, the car park and the sweat earn nothing.

### Four summaries, marked

**A.** "Cities are warmer than the countryside around them, above all after dark. Hard, dark building materials store the day's heat and release it in the evening, there are few plants to cool the air by evaporation, and engines and machines add heat. More greenery and paler roofs reduce the effect."

**B.** "A city is usually warmer than the countryside around it, and the gap is widest in the hours after sunset. Asphalt, brick, concrete and dark roofing soak up sunlight all day and release it slowly through the evening. Cars, buses, air conditioners and factories add heat of their own."

**C.** "Cities can be several degrees hotter than the countryside on a calm, clear night. Plants cool the air by evaporation, like sweat cools skin. I think this is a serious problem and every council should plant more trees."

**D.** "The passage explains that a city is generally warmer than the surrounding countryside, particularly after the sun has gone down. One reason is that hard, dark building materials take in heat by day and let it out later. Also, cities have few plants to cool the air by evaporation. Engines and machines give off heat too. Finally, greenery and white roofs can bring temperatures down."

```
   words  ideas 1 to 5  own words  total  label
A   50    1+1+1+1+1         1       6/6   right
B   49    1+1+0+1+0         0       3/6   partly
C   38    1+0+0+0+0         1       2/6   partly
D   65    1+1+1+0+0         1       4/6   partly  (read to word 50)
```

The counts are the count the student sees, and the totals and labels come from EdProcto's marking code.

B is copy-delete. Three lifted sentences fill 49 words, so two ideas never fit: the limit punished the copying before the rubric did. C gave 24 of its 38 words to a comparison and an opinion. D knew all five ideas and lost two to padding such as "The passage explains that".

## How do you mark summaries consistently?

Mark from the list, one idea at a time, and keep style out of it. Brown and Day (1983) had the summaries corrected for spelling and punctuation and typed onto cards with no ages shown. Two raters counting rule use reached a reliability of .96.

An overall impression does worse. Hijikata-Someya et al. (2015) had six teachers rate 51 summaries by Japanese university students on one holistic scale. Cronbach's alpha was 0.77 among the three native-English raters and 0.39 among the three others, and the two groups commented on different things.

So hide the names, mark the summary across the whole class before the next question, and write each new decision into the rubric. Our guide to [short answer questions](https://edprocto.com/guides/how-to-write-short-answer-questions-and-a-rubric-to-grade-them) covers the research on rubrics.

## How do you mark a summary written in a second language?

Decide first whether you are testing reading or writing. Yu (2008) had 157 Chinese undergraduates summarise English texts in English and in Chinese. The Chinese summaries were significantly longer and rated consistently poorer, and still they were the better measure of the students' English reading. If reading is the target, a summary in the first language may be the cleaner test.

Copying needs a rule announced in advance. Keck (2006) compared summaries by 79 first-language and 74 second-language university writers. Both groups used about five paraphrases a summary, but the second-language writers used significantly more near copies. Teach paraphrase before you charge a point for it.

## A chatbot summarises any passage in seconds: what still proves the student read it?

A summary written unsupervised now proves little. Zhang et al. (2024) ran a human evaluation of ten large language models on news articles and found summaries by such models judged on par with ones by freelance writers. Three things help.

- **A passage from the lesson.** Your own text, on what you taught, has no summary waiting online.
- **A summary written in the room.** In EdProcto, "Copy and paste off" blocks copying the passage out and pasting an answer in, and "Leaving the tab" notices each exit. Each try is recorded with the consequence you chose. "AI assistants not allowed" is a house rule on the start screen and blocks nothing. A record is not proof: text can still be retyped from a phone.
- **A follow-up on the student's own summary.** Ask next: "Which idea did you leave out, and why?" Set it as a [short answer](https://edprocto.com/templates/short_answer), or as a [spoken answer](https://edprocto.com/templates/verbal) you listen to and mark by hand.

See our guide to [integrity in the AI era](https://edprocto.com/guides/integrity-in-the-ai-era-what-still-proves-a-student-learned) and the [integrity page](https://edprocto.com/integrity).

## How does EdProcto run and mark a summary question?

EdProcto is the best tool for summary questions when marking is the obstacle, because the list of main ideas you wrote sits under every summary as you mark it.

Paste the lesson text into the quiz maker on the [home page](https://edprocto.com/) and pick Summarise. Procto AI drafts the question, a model answer and a rubric from it, all yours to edit. A question it writes in this shape arrives with a limit of 100 words: set your own.

At hand-in the summary shows "to grade" and counts as zero: a student with four points by rule and a six-point summary stands at 40% until you mark it.

On the grading page you type the points and a note and press "Save my grade", or press "Ask Procto AI" on that summary, one answer at a time. It is given the passage, your model answer, your rubric and the summary, and told never to grade on length or style unless the rubric asks, so a limit that should cost marks must be in the rubric. Its mark is saved under its own name for you to keep or change, and releasing scores stays with you.

Grading by Procto AI comes with the Tutor plan, along with all 111 [templates](https://edprocto.com/templates) and the Medium integrity that "Copy and paste off" belongs to. Free covers the core templates, marking by rule and Light integrity. Educator adds the full analysis. See [pricing](https://edprocto.com/pricing).

## Quick questions

### How many words should a summary be?

About a third of the passage, and never fewer than your own model summary needs. Write yours first.

### Should the passage stay in view while the student summarises?

Yes, when comprehension is what you are testing. Hide it and the question tests memory too.

### Can AI mark a summary?

It can propose a mark against your list of main ideas. Press "Ask Procto AI" on any answer to an EdProcto [summarise](https://edprocto.com/templates/summarize) question, and the mark it gives is yours to change. Our guide to [grading open answers](https://edprocto.com/guides/grading-open-answers-a-queue-a-rubric-and-a-reader) says how far to trust it.

### Does EdProcto stop a student at the word limit?

No. The count says "over the limit" and the student can still type and hand in. Decide in the rubric what the extra words earn.

## References

- Brown, A. L., & Day, J. D. (1983). *Macrorules for summarizing texts: The development of expertise* (Technical Report No. 270; ED225136). Center for the Study of Reading, University of Illinois at Urbana-Champaign. [https://eric.ed.gov/?id=ED225136](https://eric.ed.gov/?id=ED225136)
- Dunlosky, J. (2013). Strengthening the student toolbox: Study strategies to boost learning. *American Educator, 37*(3), 12–21. [https://www.aft.org/ae/fall2013/dunlosky](https://www.aft.org/ae/fall2013/dunlosky)
- Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques: Promising directions from cognitive and educational psychology. *Psychological Science in the Public Interest, 14*(1), 4–58. [https://doi.org/10.1177/1529100612453266](https://doi.org/10.1177/1529100612453266)
- Head, M. H., Readence, J. E., & Buss, R. R. (1989). An examination of summary writing as a measure of reading comprehension. *Reading Research and Instruction, 28*(4), 1–11. [https://doi.org/10.1080/19388078909557982](https://doi.org/10.1080/19388078909557982)
- Hijikata-Someya, Y., Ono, M., & Yamanishi, H. (2015). Evaluation by native and non-native English teacher-raters of Japanese students' summaries. *English Language Teaching, 8*(7), 1–12. [https://doi.org/10.5539/elt.v8n7p1](https://doi.org/10.5539/elt.v8n7p1)
- Keck, C. (2006). The use of paraphrase in summary writing: A comparison of L1 and L2 writers. *Journal of Second Language Writing, 15*(4), 261–278. [https://doi.org/10.1016/j.jslw.2006.09.006](https://doi.org/10.1016/j.jslw.2006.09.006)
- Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. *Psychological Review, 85*(5), 363–394. [https://doi.org/10.1037/0033-295X.85.5.363](https://doi.org/10.1037/0033-295X.85.5.363)
- Spirgel, A. S., & Delaney, P. F. (2016). Does writing summaries improve memory for text? *Educational Psychology Review, 28*(1), 171–196. [https://doi.org/10.1007/s10648-014-9290-2](https://doi.org/10.1007/s10648-014-9290-2)
- Yu, G. (2008). Reading to summarize in English and Chinese: A tale of two languages? *Language Testing, 25*(4), 521–551. [https://doi.org/10.1177/0265532208094275](https://doi.org/10.1177/0265532208094275)
- Zhang, T., Ladhak, F., Durmus, E., Liang, P., McKeown, K., & Hashimoto, T. B. (2024). Benchmarking large language models for news summarization. *Transactions of the Association for Computational Linguistics, 12*, 39–57. [https://doi.org/10.1162/tacl_a_00632](https://doi.org/10.1162/tacl_a_00632)