# Easy, hard, easy: question order by difficulty and test anxiety

**What order should test questions go in: easy first, hard first, shuffled or easy, hard, easy?** Question order by difficulty changes scores a little and the feel of a test more. Easy first is the safe default: pooled studies put it slightly ahead of random and hard-first orders. Easy, hard, easy keeps that start and steps down at the finish, and people prefer tests that start or end gently. Shuffle against copying; keep one order when questions build.

Published 2026-10-10 by EdProcto. https://edprocto.com/guides/easy-hard-easy-ordering-questions-to-lower-test-anxiety

The same twelve questions can be set easiest first, hardest first, shuffled, or easy, hard, easy. Researchers call the choice item order.

Question order by difficulty moves scores a little, and in several studies not at all. It moves how a test feels, and how it is remembered, more dependably.

EdProcto, which publishes this guide, builds the easy-hard-easy order as its [Sandwich](https://edprocto.com/templates/sandwich) template: you label each question and its server puts the hard ones in the middle. Each source below was read on 10 October 2026.

## Which order should test questions go in?

Start with easy questions unless you have a reason not to, then pick the rest of the order by what the test is for.

| Order | What the studies we read say | Use it for |
| --- | --- | --- |
| Easy to hard | Slightly higher scores when pooled; the fewest people giving up | Most tests |
| Hard to easy | Lower scores in one study; the most anxiety in one course of another | Nothing we can name |
| Shuffled | A small average cost when studies are pooled | Stand-alone questions, students side by side |
| Easy, hard, easy | Untested head to head; rests on how tests are remembered | Low-stakes quizzes, in EdProcto's [Sandwich](https://edprocto.com/templates/sandwich) |
| As written | Questions after related ones were answered better | Questions that build |

## What does research say about question order and scores?

It says an easy start does at least as well as any other order, and that its advantage is small.

### Classroom tests

Aamodt and McShane (1992) pooled the experiments then published. By their abstract, tests running from easy to hard came out slightly easier than randomly ordered tests (d = 0.11) and than tests that began hard (d = 0.22), effects the authors call small.

Single studies disagree. Hambleton and Traub (1974) found fewer right answers on a mathematics test arranged from difficult to easy than on its reverse. Plake et al. (1981) gave 97 undergraduates a 48-question mathematics exam in three orders and found no significant difference, adding that their design may have lacked the power to find one. Vander Schee (2013) found no significant effect of difficulty order across 1,335 marketing exams.

### Large testing programmes

Anaya et al. (2022) ran a field experiment with an online teaching platform, assigning about 19,000 participants at random to different orders of difficulty. Easiest to hardest gave the lowest probability of abandoning the test and the most correct answers, and the authors report similar results from the differences between test booklets in the international PISA assessments of 2009, 2012 and 2015. Their abstract gives no effect size.

Reviewing the field with the stability of a question's statistics in mind, Leary and Dorans (1985) found evidence of context effects, and none strong enough to overturn practice that depends on that stability.

Our reading: the difference can vanish in one classroom, and it grows where people are free to walk away.

## Does question order lower test anxiety?

No study we read shows that an easy-hard-easy order does, and the evidence on order and anxiety is thin. Of the three studies below, two found that order changed stress or anxiety and one that it changed neither scores nor attitudes.

Hambleton and Traub (1974) report that item order affected the stress their test produced. Tippets and Benson (1989) gave 128 graduate students in two courses a real classroom exam in one of three arrangements, without time pressure. Anxiety during the exam differed by arrangement in both courses: highest on the hard-to-easy paper in one, on the random paper in the other. Plake et al. (1981) took several measures of anxiety with their exam and found no significant effect of order on scores or on how students saw the test.

### What we could not confirm

We read each study in its abstract, as the publisher, PubMed, a university repository or an open index holds it; no full text was open to us. No abstract gives the size of an effect on anxiety, so we give none. No experiment we found set easy, hard, easy against the other orders: its case is assembled, by us, from studies of starts and of ends. Nothing here is advice about anxiety as a health condition.

## Does the end of a test colour how students remember it?

Yes, and so does the start: people prefer a test that begins or ends on easier questions, even when that makes it longer.

The idea comes from Kahneman et al. (1993). Volunteers held a hand in water at 14 °C for 60 seconds; on another trial they did the same, then kept it in for 30 seconds more while the water warmed to 15 °C. A significant majority chose to repeat the longer trial. An unpleasant episode, the authors concluded, is remembered mostly by its worst and its last moments: the peak-end rule.

Finn and Miele (2016) ran the design with tests, in five experiments. People sat a test of 30 difficult maths problems, and a longer one of 30 difficult problems plus 10 moderately difficult ones. They preferred the longer test and chose it for next time. Later experiments moved the moderate problems to the start or the middle, and the authors conclude that test takers favour tests that start or end on a high note.

The start shapes judgement as well: in Weinstein and Roediger (2010), people believed they had answered more general knowledge questions correctly when the questions ran from easiest to hardest than when the same questions were shuffled or reversed.

One study with children loosens the word "end". The University of Maryland College of Education (2025) describes work by Finn, Miele and Wigfield with 281 third graders and 289 sixth graders. Both ages preferred ten difficult problems with five easier ones added to the ten alone, and where the easier five sat, beginning, middle or end, made little or no difference. We read the university's account, not the paper.

These studies measure which test people would sit again and how well they think they did, not anxiety and not learning.

## Is it fair when students get the questions in different orders?

It is fair enough when each order is drawn at random and every question stands alone; when questions build, one order for everyone is fairer.

Hambleton and Traub (1974) took their result as evidence against giving examinees different orders to make copying harder. Vander Schee (2013), finding no significant effect, concluded that teachers of introductory marketing courses can randomise in comfort. A pooled d of 0.11 allows both readings: across a class a shuffle costs little, and the student who draws three hard questions first pays more than the average (our inference).

### When a fixed order is right

Keep one order when questions lean on each other, as in a [case study](https://edprocto.com/templates/case) or a [scaffolded sequence](https://edprocto.com/guides/scaffolded-questions-one-problem-in-steps). Haasnoot et al. (2026) studied 16,127 multiple-choice answers across 12 exams in microeconomics at a Dutch university: a question was answered correctly more often when the one before it came from the same lecture or the previous one.

### When a shuffle is fairer

Shuffle the questions when students can see each other's screens and copying is the larger threat. Our guide to [the unit test online](https://edprocto.com/guides/the-unit-test-online-settings-that-keep-it-fair) weighs shuffling against copying, with the settings.

## How do you judge a question's difficulty before anyone has sat it?

Judge it by what the student has to do, get a second opinion, and expect to be wrong about a few.

Impara and Plake (1998) asked 26 classroom teachers to estimate how their own students would do on each question of a district science test: accuracy was not high, even for the class as a whole. Among 48 trainee teachers, Nathan and Petrosino (2003) found those with more advanced mathematics likelier to treat equations as a prerequisite for word and story problems, a view at odds with how students actually perform.

So label in four steps, which are ours.

1. **Sit the quiz cold** and note where you slowed down.
2. **Count what each question asks for.** One fact recalled is easy. One method applied to new numbers is medium. Two methods chained, or a familiar trap, is hard.
3. **Get a second opinion**: a colleague, or last year's results.
4. **Label a third of the quiz hard at most.**

### Correcting the order from results

After the first sitting, let the class overrule you. On EdProcto's results page, "Each question" shows the share who got each one right under "Got it", on every plan, and its "Heat map" colours each question in one of three bands: "under 50% got it", "50 to 70%" and "over 70%". Relabel those hard, medium and easy before the next class sits the quiz; the mapping is our working rule. Where more than half missed a question you called easy, read its key first. [Item analysis](https://edprocto.com/guides/item-analysis-for-teachers-what-the-numbers-mean) explains how rough these figures are in one class.

## What does one quiz look like in three orders? A worked example

It looks like three different papers. The quiz is our example, on percentages for Year 8, and the labels are a teacher's guess.

| Label | Questions, numbered as built |
| --- | --- |
| Easy | 1: 50% of 80. 2: 0.25 as a percentage. 3: 10% of 250. 4: 3/4 as a percentage. |
| Medium | 5: 15% of 60. 6: increase 40 by 25%. 7: 12 out of 48 as a percentage. 8: a £30 shirt at 20% off. |
| Hard | 9: a price rises 20% to £60; find the original. 10: 200 up 10%, then down 10%. 11: 35% of a number is 21; find it. 12: £500 at 4% compound interest for two years. |

Answers: 40, 25%, 25, 75%, 9, 50, 25%, £24, £50, 198, 60 and £540.80.

| Arrangement | The paper | What it risks |
| --- | --- | --- |
| Easy to hard | 1 2 3 4 5 6 7 8 9 10 11 12 | The paper ends at its worst, and slow students meet the hardest four with the least time |
| Shuffled | One draw: 5 4 11 10 1 12 9 6 3 8 7 2 | Two hard questions in this student's first four; another draw opened on 8 and 9 |
| Easy, hard, easy | 1 2 5 6 9 10 11 12 7 8 3 4 | A student who stalls on 9 to 12 may run short of time for 7, 8, 3 and 4, marks that were in reach |

The last two rows came from EdProcto's own ordering code, run on these labels. The third is the Sandwich: easy, medium, hard, medium, easy. The paper climbs to its peak and steps down again, as the longer test in Finn and Miele (2016) stepped down.

## How does EdProcto build the easy-hard-easy order?

EdProcto is the best tool for an easy-hard-easy quiz because its server makes the shape from your labels by rule, the same for every student, and holds it whatever else you set.

- **Building.** Choose Sandwich among the Whole-quiz structures. A new one starts with a single multiple-choice question, and "Add a question" asks "Which shape?", so each question takes the shape that suits it.
- **Labelling.** Every marked question carries "Difficulty": "Easy", "Medium" or "Hard". A new question starts at "Medium".
- **The rule.** When a student begins, the server halves the Easy questions and the Medium ones, each kept in the order you built them. The first half of the Easy ones opens the paper, then the first half of the Medium; every Hard question follows; the other Medium ones come next and the other Easy ones close the paper. An odd one goes to the opening. With two labels only, Easy and Hard, the Easy questions are shared between the two ends.
- **Checking.** "Preview as student" shows the paper as a student meets it. The results page and "Paper" number the questions as they stand in the builder, so press "Move up" and "Move down" until the builder matches the paper.
- **Settings.** On the settings page "Shuffle the questions for each student" is greyed and reads "Not applicable on this template · the order of the questions is the structure". The labels decide the order whichever preset you press, and "Shuffle the options" stays yours to set. The Exam [preset](https://edprocto.com/guides/quiz-settings-explained-practice-standard-and-exam-presets) switches off "Allow going back to change an answer" and "Allow skipping a question": on a timed Sandwich switch both on again, so a student can leave a hard question and return to it.

Where the score counts, add the [integrity](https://edprocto.com/integrity) measures the stakes call for. They raise the cost of copying without ruling it out, and the score stays yours to release.

Free covers ten quizzes a month, the core templates and the results question by question, "Got it" included. All 111 [templates](https://edprocto.com/templates) come with Tutor, and the full analysis with Educator. [Pricing](https://edprocto.com/pricing) has the figures.

## Quick questions

### Does question order reduce test anxiety?

The evidence is thin. Tippets and Benson (1989) found anxiety highest on a hard-to-easy paper in one course and on a random one in another; Plake et al. (1981) found no effect of order on scores or attitudes.

### What is the peak-end rule?

The finding that an unpleasant episode is remembered mostly by its worst and its last moments (Kahneman et al., 1993). On tests, people preferred a longer paper that started or ended on easier problems (Finn & Miele, 2016).

### Is it fair to shuffle questions for each student?

Yes, when the questions stand alone: the average cost is small (Aamodt & McShane, 1992) and each order is drawn at random. Keep one order when questions build.

### How does EdProcto order a Sandwich quiz?

By the "Difficulty" label on each question: Easy questions open and close the paper, Medium ones stand either side of the middle and Hard ones sit in it, the same for every student. The [Sandwich template](https://edprocto.com/templates/sandwich) page shows the shape.

## References

- Aamodt, M. G., & McShane, T. (1992). A meta-analytic investigation of the effect of various test item characteristics on test scores and test completion times. *Public Personnel Management, 21*(2), 151–160. [https://doi.org/10.1177/009102609202100203](https://doi.org/10.1177/009102609202100203)
- Anaya, L., Iriberri, N., Rey-Biel, P., & Zamarro, G. (2022). Understanding performance in test taking: The role of question difficulty order. *Economics of Education Review, 90*, Article 102293. [https://doi.org/10.1016/j.econedurev.2022.102293](https://doi.org/10.1016/j.econedurev.2022.102293)
- Finn, B., & Miele, D. B. (2016). Hitting a high note on math tests: Remembered success influences test preferences. *Journal of Experimental Psychology: Learning, Memory, and Cognition, 42*(1), 17–38. [https://doi.org/10.1037/xlm0000150](https://doi.org/10.1037/xlm0000150)
- Haasnoot, C. W., van Krevel, C., & de Vaal, A. (2026). Ordered or scrambled: How the forward sequencing of multiple choice questions affects test item scores. *Assessment & Evaluation in Higher Education, 51*(2), 247–262. [https://doi.org/10.1080/02602938.2025.2568187](https://doi.org/10.1080/02602938.2025.2568187)
- Hambleton, R. K., & Traub, R. E. (1974). The effects of item order on test performance and stress. *The Journal of Experimental Education, 43*(1), 40–46. [https://doi.org/10.1080/00220973.1974.10806302](https://doi.org/10.1080/00220973.1974.10806302)
- Impara, J. C., & Plake, B. S. (1998). Teachers' ability to estimate item difficulty: A test of the assumptions in the Angoff standard setting method. *Journal of Educational Measurement, 35*(1), 69–81. [https://doi.org/10.1111/j.1745-3984.1998.tb00528.x](https://doi.org/10.1111/j.1745-3984.1998.tb00528.x)
- Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). When more pain is preferred to less: Adding a better end. *Psychological Science, 4*(6), 401–405. [https://doi.org/10.1111/j.1467-9280.1993.tb00589.x](https://doi.org/10.1111/j.1467-9280.1993.tb00589.x)
- Leary, L. F., & Dorans, N. J. (1985). Implications for altering the context in which test items appear: A historical perspective on an immediate concern. *Review of Educational Research, 55*(3), 387–413. [https://doi.org/10.3102/00346543055003387](https://doi.org/10.3102/00346543055003387)
- Nathan, M. J., & Petrosino, A. (2003). Expert blind spot among preservice teachers. *American Educational Research Journal, 40*(4), 905–928. [https://doi.org/10.3102/00028312040004905](https://doi.org/10.3102/00028312040004905)
- Plake, B. S., Thompson, P. A., & Lowry, S. (1981). Effect of item arrangement, knowledge of arrangement, and test anxiety on two scoring methods. *The Journal of Experimental Education, 49*(4), 214–219. [https://doi.org/10.1080/00220973.1981.11011786](https://doi.org/10.1080/00220973.1981.11011786)
- Tippets, E., & Benson, J. (1989). The effect of item arrangement on test anxiety. *Applied Measurement in Education, 2*(4), 289–296. [https://doi.org/10.1207/s15324818ame0204_2](https://doi.org/10.1207/s15324818ame0204_2)
- University of Maryland College of Education. (2025, March 25). *Remembered success: How adding easier math problems helps students choose to tackle harder ones*. [https://education.umd.edu/node/43688](https://education.umd.edu/node/43688)
- Vander Schee, B. A. (2013). Test item order, level of difficulty, and student performance in marketing education. *Journal of Education for Business, 88*(1), 36–42. [https://doi.org/10.1080/08832323.2011.633581](https://doi.org/10.1080/08832323.2011.633581)
- Weinstein, Y., & Roediger, H. L., III. (2010). Retrospective bias in test performance: Providing easy items at the beginning of a test makes students believe they did better on it. *Memory & Cognition, 38*(3), 366–376. [https://doi.org/10.3758/MC.38.3.366](https://doi.org/10.3758/MC.38.3.366)