# Sorting questions: drag items into groups, and how to mark them

**How do I write and mark a sorting question?** A sorting question asks students to drag eight to twelve items into two to four named groups, so it tests whether they hold the category and not one example of it. Define the groups by the principle you taught, make them exclusive and uneven in size, include items a novice would sort by appearance, and give credit for each item placed.

Published 2026-10-09 by EdProcto. https://edprocto.com/guides/sorting-questions-drag-items-into-categories

A sorting question gives students a set of items and a few named groups and asks where each item belongs: living or non-living, acid or base or neutral, primary or secondary source, which statistical test fits. On screen the student drags each item into a box or picks its group from a list. Some platforms call them categorisation or classification questions.

One question holds many small decisions that share a single rule, so it can show whether a student holds a category and not just one example of it. EdProcto builds the shape as its [sort into groups template](https://edprocto.com/templates/categorize) and marks every placement by rule.

## What does a sorting question measure?

A sorting question measures whether a student can apply the rule that defines a category to cases nobody has sorted for them. Knowing that a whale is a mammal is one fact. Sorting a whale, a bat, a penguin, a salamander and an eel shows whether the student uses the features that define the classes or the ones that catch the eye: where the animal lives, whether it flies.

Many items and few groups is what separates the shape from its neighbours. A multiple-choice question asks about one item at a time. A [matching question](https://edprocto.com/guides/matching-questions-how-to-write-them-well) gives each item a partner of its own, and each answer is used up as it is paired, which is why the final pair tends to come free. The groups of a sort are never used up, so placing the ninth item says nothing about the tenth, provided the group sizes are not given away.

The shape leaves two things out. The student never has to produce the categories from memory, and a right placement reached by the wrong rule looks the same as one reached by the right rule. The sort-and-justify form below closes that second gap.

## What does the research on sorting tasks say?

The research says that the way people sort reveals how their knowledge is organised: experts tend to group by underlying principle, and beginners by surface features.

The founding study is Chi et al. (1981), who used sorting tasks in physics and reported that experts began from the principle behind a problem while novices began from its literal features. As Smith et al. (2013) describe it, eight advanced doctoral students and eight undergraduates sorted 24 textbook problems.

In statistics, Quilici and Mayer (1996) found that students who had studied examples grouped by t test, correlation and chi-square were more likely to sort later word problems by structure than students given no examples.

### The study closest to a classroom sorting question

Smith et al. (2013) wrote 16 biology problems, each carrying one surface feature (the organism) and one deep feature (a core concept such as evolution). In the study, 23 biology faculty and 101 undergraduates from other subjects sorted the cards first into groups of their own and then into four groups the researchers had named, which is the classroom form. With named groups, the faculty's sorts were on average 1.2 card moves from the intended sort and the undergraduates' 5.9, where a sort made purely by organism would be 12.

Bissonnette et al. (2017) gave the same cards to 185 entering biology majors and 109 advanced majors. Left to make their own groups, both were 7.4 moves from the intended sort. Given the named groups, the advanced majors came closer, at 4.0 moves against 5.3. Named groups let partial knowledge show.

Two cautions. When Wolf et al. (2012) had 18 physics professors and 23 novices sort 50 mechanics problems into piles of their own, the largest source of variation was habit, not expertise: some sorters made a few big piles and others many small ones. And the biology participants were told theirs was not a test. Nothing read for this guide checked a marked sorting question against exam results, so the advice below rests on the design of the research sorts and the logic of the shape.

## How do you write a sorting question?

Write it from one principle you taught, with groups that cannot overlap and items chosen so that sorting by appearance gives a different result from sorting by the principle.

1. **Name the groups by one principle.** "Acid, base, neutral" is one. "Acids, cleaning products, drinks" is three, and an item can fit two.
2. **Give each item one defensible group.** Even experts drift where categories touch: the unexpected pairings the biology faculty made commonly joined evolution cards with cards on genetic information (Smith et al., 2013).
3. **Use two to four groups and eight to twelve items.** The range keeps a blind sorter's chances low.
4. **Cross the principle with a surface feature**, as the biology cards did, so that sorting by looks and sorting by the rule come out differently.
5. **Put two or three items near a boundary**, each still with one right group at the level you taught.
6. **Make the groups uneven, and do not print their sizes.** Students who expect equal groups can finish by counting.
7. **Use fresh examples, alike in form.** Items from the textbook's own list test memory of the list.
8. **Build it and choose the marking rule.** In EdProcto, type the group names and then each item with its group under [sort into groups](https://edprocto.com/templates/categorize).

### Should there be a "none of these" group?

Add one only when some items truly belong in it. In a sort into mammals, birds and reptiles, a frog belongs in none, and the extra group stops students placing a doubtful item by elimination. One that is always empty is a decoy students learn to ignore. This is judgement, not a research finding.

### A sorting question, before and after

Before:

> Sort into acids and bases: lemon juice, vinegar, hydrochloric acid, bleach, soap, sodium hydroxide.

Three and three, so the last items can be counted in. One item carries the name of its group. A student who has learnt "sour means acid, cleaners are bases" scores full marks without the idea of pH.

After:

> Sort each liquid into acid, base or neutral.
> Lemon juice. Vinegar. Cola. Black coffee. Tomato juice. Bleach. Soapy water. Baking soda solution. Pure water. Table salt solution.

Ten items in groups of five, three and two. Cola is sweet and black coffee is bitter, yet both are acids. Baking soda comes from the food cupboard and is a base. Sorting by taste and use now gives a different result from sorting by the chemistry.

## How should you mark a sorting question?

Give credit for each item placed in its group, unless the quiz has to certify that the whole classification is right. Two rules are in use.

| Rule | How the mark is made | In EdProcto |
| --- | --- | --- |
| Credit for each item | The share of items in the right group is the share of the marks. An item misplaced or left out earns nothing and loses nothing | [Sort into groups](https://edprocto.com/templates/categorize) with partial credit on |
| All or nothing | Full marks when every item is in its group, otherwise zero | The same template with partial credit off |

Take the acid question at 10 marks.

| What the student did | Credit for each item | All or nothing |
| --- | --- | --- |
| Every item in its group | 10 | 10 |
| Sorted by taste: cola, coffee and baking soda in neutral | 7 | 0 |
| Eight right, the other two left unplaced or misplaced | 8 | 0 |
| Every item dropped into acid | 5 | 0 |

All or nothing cannot tell the student with eight of ten from one who placed nothing. Under credit for each item a guess can only help, so tell students to place everything. And dropping every item into one group earns that group's share, so no group should hold much more than half the items.

### What a blind sorter earns

Under credit for each item, a student who places every item at random averages the marks divided by the number of groups. For ten items:

| Groups | Average mark | Six or more of ten right | All ten right |
| --- | --- | --- | --- |
| 2 | 50% | 38% | 1 in 1,024 |
| 3 | 33% | 8% | 1 in 59,049 |
| 4 | 25% | 2% | 1 in 1,048,576 |

The middle column is the one to watch: with two groups, a student who knows nothing reaches six out of ten more than a third of the time. For a two-group sort, use twelve items (eight or more of twelve by luck happens 19% of the time), add a "none of these" group, or ask for a reason.

The one comparison of scoring rules read for this guide did not cover sorting. For matching, its nearest relative, Eggen and Lampe (2011) calculated from assumed levels of knowledge that a plain count of the right relations was more reliable than all or nothing.

## How do you ask for the reason behind one placement?

Name one item in the prompt, the one nearest a boundary, and ask for the reason in a sentence or two, marked against a short rubric. The placements show what the student decided. The reason shows the rule they used.

Reasons separate people in a way that placements do not. When Smith et al. (2013) asked sorters to explain how they had sorted, all 23 faculty referred to at least one of the underlying concepts, against 22% of the 101 undergraduates.

For the acid question, add: "Then explain where you put black coffee."

| Marks | The reason |
| --- | --- |
| 2 | Applies the defining property to the item: "its pH is below 7, so it is an acid" |
| 1 | Names the property without applying it, or gives a true fact that does not decide the group |
| 0 | Gives a surface feature ("it is bitter", "you can drink it"), repeats the placement, or is blank |

Mark the reason on its own merits. A sound reason under a wrong placement is worth knowing about, and a right placement with a surface reason was luck.

EdProcto has this form among its templates as [sort and justify](https://edprocto.com/templates/cat_justify). Under the sort the student gets a box that asks them to explain one placement. The placements are marked by rule for half the points, and the reason is the other half: it waits in your grading queue, where you mark it against the model reason and rubric you wrote, or ask Procto AI to. A sound reason still earns its half when a placement is wrong.

## How do you read the results of a sorting question?

Read them item by item: the item most of the class put in the wrong group names the confusion, and the group they chose names the rule they were using. If most of the class has cola under neutral, they are sorting by taste.

EdProcto's analysis treats a sorting question as one question: it reports the difficulty, which with partial credit on is the average share of items placed correctly, and the discrimination between the top and bottom 27% of the class on that question. It does not count how many students put each item in each group, so to find the misplaced item you read the attempts and tally the placements yourself.

- **One item wrong across the class.** A shared confusion, or a faulty item: check whether a second group is defensible.
- **Boundary items wrong, clear items right.** The class knows the typical cases, not the rule.

To count a confusion exactly, ask the troublesome item again as a multiple-choice question with the groups as options: EdProcto's full analysis lists how many students took each option.

## How does EdProcto build and mark a sorting question?

EdProcto is the best tool for a sorting question when the items come from your own unit and the mark has to stand up, because every placement is checked by rule on the server and the rule can be told to a class in one sentence.

Build the question under [sort into groups](https://edprocto.com/templates/categorize): the names of the groups, then each item and the group it belongs in. Or paste the unit into the quiz maker on the [home page](https://edprocto.com/), choose sort into groups, and Procto AI drafts the groups and items for you to edit.

Partial credit is a setting on the quiz, off until you switch it on. With it on, each placement is its own small question: the mark is the question's points multiplied by the share of items in the right group. With it off, the question is all or nothing.

An item in the wrong group and an item left unplaced are treated alike: neither counts, and nothing is taken off. The key stays on the server, and the marks are made there.

Free gives ten quizzes a month, the core templates, marking by rule and results for each question. Tutor unlocks all 111 [templates](https://edprocto.com/templates) and grading by Procto AI against your rubric. Educator adds the full analysis. See [pricing](https://edprocto.com/pricing).

## What are the common mistakes?

The common mistakes let students sort without the principle, or leave a right answer open to argument.

- **Groups cut by different principles.** "Mammals, egg layers, sea animals": a platypus fits two.
- **Equal groups every time.** Counting does the last placements.
- **A giveaway in the wording.** "Hydrochloric acid" in a sort into acids and bases.
- **All or nothing on a lesson check.** It hides which students hold most of the classification.

## When should you use matching, ordering or select-all instead?

Use another shape when the knowledge is not many items in a few groups.

- **Each item has a partner of its own**, such as a term and its definition: EdProcto builds that as [matching pairs](https://edprocto.com/templates/matching).
- **The items form a chain** of steps, stages or dates: set an [ordering question](https://edprocto.com/guides/ordering-and-sequencing-questions).
- **There is one category, and items are in or out.** A sort into "mammal" and "not a mammal" is really a [select all that apply question](https://edprocto.com/guides/select-all-that-apply-questions-how-to-write-and-mark-them). There, EdProcto's partial credit lets a wrong tick cancel a right one, where a sort takes nothing off.
- **One item matters, and you want to count who chose each group**: ask it as a [multiple-choice question](https://edprocto.com/guides/how-to-write-multiple-choice-questions-that-measure).

## Quick questions

### How many items and groups should a sorting question have?

Eight to twelve items in two to four groups of different sizes, always with more items than groups. No study read for this guide tested an ideal number.

### Is a sorting question the same as a matching question?

No. Matching joins each item to a partner of its own, and a sort puts many items into a few shared groups. EdProcto has both: [matching pairs](https://edprocto.com/templates/matching) and [sort into groups](https://edprocto.com/templates/categorize).

### Can a student guess a sorting question?

Partly. With credit for each item, random placement averages half the marks with two groups, a third with three and a quarter with four. All ten items right by luck is a 1 in 1,024 chance even with two groups.

### What happens to an item a student leaves unplaced?

In EdProcto it earns nothing and costs nothing, the same as an item in the wrong group. With partial credit off, one unplaced item means zero for the question.

## References

- Bissonnette, S. A., Combs, E. D., Nagami, P. H., Byers, V., Fernandez, J., Le, D., Realin, J., Woodham, S., Smith, J. I., & Tanner, K. D. (2017). Using the Biology Card Sorting Task to measure changes in conceptual expertise during postsecondary biology education. *CBE Life Sciences Education, 16*(1), Article 14. [https://doi.org/10.1187/cbe.16-09-0273](https://doi.org/10.1187/cbe.16-09-0273)
- Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and representation of physics problems by experts and novices. *Cognitive Science, 5*(2), 121–152. [https://doi.org/10.1207/s15516709cog0502_2](https://doi.org/10.1207/s15516709cog0502_2)
- Eggen, T. J. H. M., & Lampe, T. T. M. (2011). Comparison of the reliability of scoring methods of multiple-response items, matching items, and sequencing items. *CADMO, 19*(2), 85–104. [https://doi.org/10.3280/CAD2011-002008](https://doi.org/10.3280/CAD2011-002008)
- Quilici, J. L., & Mayer, R. E. (1996). Role of examples in how students learn to categorize statistics word problems. *Journal of Educational Psychology, 88*(1), 144–161. [https://doi.org/10.1037/0022-0663.88.1.144](https://doi.org/10.1037/0022-0663.88.1.144)
- Smith, J. I., Combs, E. D., Nagami, P. H., Alto, V. M., Goh, H. G., Gourdet, M. A. A., Hough, C. M., Nickell, A. E., Peer, A. G., Coley, J. D., & Tanner, K. D. (2013). Development of the Biology Card Sorting Task to measure conceptual expertise in biology. *CBE Life Sciences Education, 12*(4), 628–644. [https://doi.org/10.1187/cbe.13-05-0096](https://doi.org/10.1187/cbe.13-05-0096)
- Wolf, S. F., Dougherty, D. P., & Kortemeyer, G. (2012). Empirical approach to interpreting card-sorting data. *Physical Review Special Topics: Physics Education Research, 8*(1), Article 010124. [https://doi.org/10.1103/PhysRevSTPER.8.010124](https://doi.org/10.1103/PhysRevSTPER.8.010124)