How to Build Comparable Versions of the Same Exam
Create more comparable exam versions with a shared blueprint, item pairing sheet and review checks, without claiming statistical equivalence.
Two versions of an exam are not comparable just because they have the same number of questions. They should sample the same learning outcomes, similar cognitive demands, formats, marks and likely difficulty. For an ordinary classroom exam, this means disciplined blueprinting and review. For high-stakes score comparability, formal equating or specialist measurement work may be needed; a teacher-made checklist cannot establish statistical equivalence.
ETS defines alternate-forms reliability as consistency across different editions of a test covering the same kinds of knowledge and skills at the same difficulty level. Its reliability primer explains why comparability is a measurement issue, not merely a formatting issue.
Start with a shared specification
Write one blueprint for both forms. Suppose a 40-mark science exam has four sections:
| Section | Purpose | Format | Marks in Form A | Marks in Form B |
|---|---|---|---|---|
| 1 | Core concepts | 10 multiple-choice items | 10 | 10 |
| 2 | Data interpretation | 2 short responses | 10 | 10 |
| 3 | Application | 2 structured problems | 12 | 12 |
| 4 | Explanation | 1 extended response | 8 | 8 |
Matching these totals is only a starting point. If Form A’s data task uses a simple labelled table and Form B’s uses a dense unfamiliar graph, the workload and challenge may differ. Review the actual items, not just the spreadsheet.
Build a pairing sheet
For each item or task in Form A, identify the intended match in Form B by outcome, topic, demand, format, marks and estimated time. Example:
| Form A task | Form B task | Shared target | Check |
|---|---|---|---|
| Interpret a rainfall table | Interpret a temperature table | Identify trend and support it with data | Both have similar number of data points and reading burden |
| Explain a circuit failure | Explain a different circuit failure | Apply series-circuit reasoning | Both require the same number of reasoning steps |
The examples are illustrative, not validated equivalent items. Ask subject teachers to compare the cognitive work required, not just the surface topic.
Review difficulty and timing
- Have reviewers complete both forms without seeing the answer keys.
- Record time taken per section and note unusually confusing wording.
- Compare the difficulty mix using the same difficulty-level convention for both forms.
- Check that any images, source passages and calculations demand similar effort.
- Review each form’s mark scheme and accepted alternative responses.
If the exam will make a consequential decision, consider a small pilot and appropriate expert advice. Do not claim the forms are statistically equivalent from a visual review alone.
Keep version control unambiguous
Label every paper and key with Form A or Form B, date, version and approval status. Do not reuse a question in a way that gives one group prior access to the other’s answers. Keep the final PDF and answer key paired. If you regenerate a form, check that the new output still meets the blueprint.
Our exam integrity guide discusses why multiple versions may reduce answer sharing. The separate issue addressed here is fairness of content and challenge across versions. A question paper generation workflow can help draw from a reviewed bank, but human review remains essential.
When to stop and seek measurement support
If two forms will be used to compare cohorts, determine certification outcomes or set high-stakes cut scores, the implications are larger than a classroom practice test. ETS research on score-scale inconsistency notes that equating relies on assumptions and data quality. Engage a qualified assessment specialist rather than presenting matched blueprints as proof of equal scores.
Bottom line: a shared blueprint and item-level review make alternate forms more defensible. They do not, by themselves, prove statistical interchangeability.
Randomising order is not the same as writing another form
Shuffling the same questions changes their order, but the content is still identical. It may help with answer copying in some settings, but it does not create a new set of items. Two forms contain different questions built to a common specification. Keep that distinction clear when explaining the approach to staff or students.
Also consider whether groups will sit at different times. Even a well-matched form may not solve content leakage if one group shares the topic or key with another. Exam security, supervision and publication of answers need a separate plan. For a low-stakes practice quiz, exact form comparability may matter less than giving each learner useful feedback; for a graded common assessment, it matters more.