Question Bank
The question bank is the centre of Next Exams. Questions live independently of any exam, filed against a hierarchical subject tree, and exams reference them. Write a question once and it is available to every exam, every pool rule and every report.
Questions can also be written inside an exam, in the Exam Studio. Those are kept in the bank too unless the author switches Keep in bank off, in which case they belong to that exam alone and are not listed here.
A question is edited on a page of its own. If you try to leave that page with changes you have not saved, it asks first; closing or reloading the browser tab asks too.
This page covers the sixteen question types, how each one is stored and graded, and the authoring features layered on top of them — subjects, tags, custom fields, maths, media and versioning.

Question types
Sixteen types are available. Fourteen grade automatically; essay and speaking are marked by a person.
| Type | What the examinee does | Graded |
|---|---|---|
| Multiple choice (single) | Picks exactly one option | Automatically |
| Multiple choice (multiple) | Picks any number of options | Automatically, with optional partial credit |
| True / false | Picks one of two fixed options | Automatically |
| Short answer | Types a word, phrase or number | Automatically, against an accepted-answer list |
| Essay | Writes free text | By a grader |
| Ordering | Arranges items into the correct sequence | Automatically, per position |
| Matching | Pairs each item on the left with one on the right | Automatically, per pair |
| Extended matching (EMQ) | Answers several stems from one shared option bank | Automatically, per stem |
| Cloze | Fills blanks embedded in a passage | Automatically, per blank |
| Image choice | Picks the right picture, or several, from a set of images | Automatically, with optional partial credit |
| Range | Sets a number, or an interval, on a sliding scale | Automatically, within a tolerance |
| Hotspot | Drops a pin on the right place in an image or map | Automatically |
| Image marking | Draws over the right region of an image | Automatically, by how well the drawing fits |
| Graph | Plots points on a coordinate grid | Automatically, per point |
| Puzzle | Puts the shuffled pieces of a picture back in place | Automatically, per piece |
| Speaking (recorded) | Records a spoken answer to a role-play card | By a grader, with a rubric. See Speaking questions |
Ordering, matching, EMQ and cloze are grouped as advanced types, and image choice, range, hotspot, image marking, graph and puzzle as visual types; speaking belongs to the Speaking app. Each can be switched off in the App Manager. Doing so stops authors creating new questions of those types; existing ones keep rendering and grading normally, so switching an app off can never break a live exam.
Visual question types
Six types are answered by looking and pointing rather than by reading and typing.

| Type | You set | The learner |
|---|---|---|
| Image choice | Two or more pictures from the media library, each with a description, and which are right | Chooses one picture, or several |
| Range | A scale (from, to, step, optional unit), the correct number or interval, and a tolerance | Drags a slider or types the number |
| Hotspot | A picture, and areas drawn on it — at least one right and one wrong — each with a name | Drops one pin on the picture |
| Image marking | A picture, the region that is right, and how closely a drawing must fit it | Draws over the region |
| Graph | A grid (its range and step), the points to plot, and a tolerance | Clicks the grid to plot each point, or types it as x, y |
| Puzzle | A picture and how many pieces across and down (2 to 6) | Puts the shuffled pieces back |
Drawing areas. The hotspot and image marking editors have four tools above the picture:
| Tool | How |
|---|---|
| Select | Click an area to select it, drag it to move it, or drag a handle to reshape it. With an area selected, the arrow keys move it and Delete removes it |
| Rectangle | Drag on the picture. Hold Shift for a square |
| Circle | Drag from the centre outwards |
| Polygon | Click to place each corner (up to 40). Click the first corner again, double-click or press Enter to finish; Esc cancels |
Every area is also a row under the picture, with its name, whether it is right or wrong, and a Position and size button that opens its numbers — a polygon's corners can be typed, added and removed there. That is how you make or adjust an area from the keyboard. Areas are stored as fractions of the picture, so they are the same on a phone as on a wide screen.
Range, graph and puzzle. The range editor shows the scale as a slider: drag the answer onto it, and the accepted band is drawn around it. The graph editor draws the grid as you set it; click it to plot a point and drag a point to move it; with a point selected, the arrow keys move it and Delete removes it. The puzzle editor shows the pieces shuffled the way a learner first sees them, and can show the finished picture.
Image marking's list of regions. Under the picture you name the region that is right and at least one that is wrong. A learner who cannot draw chooses from these names instead. Exactly one is right, so they are a single group of radio buttons.
How image marking is graded. A drawing passes when it covers at least a set share of the region (70% by default) and no more than a set share of the drawing lies outside it (30% by default). The second rule is what stops a learner passing by painting the whole picture.
Every visual type can be answered without a pointer. This is not an afterthought: the control that works from the keyboard, with a screen reader and without JavaScript is the answer, and the slider, pin, brush, grid and board are drawn over it.
| Type | Without a pointer |
|---|---|
| Image choice | Ordinary radio buttons or checkboxes; each picture's description is read out |
| Range | A number box |
| Hotspot | Arrow keys move the pin, or the place is chosen by name from a list |
| Image marking | Arrow keys and Space mark the picture, or the region is chosen by name from a list |
| Graph | Each point is typed as x, y |
| Puzzle | A position is chosen for each piece; on the board, Enter picks a piece up and puts it down |
This is why a hotspot needs at least one wrong area and an image marking question needs a list of named regions: they are what the learner chooses between when choosing by name.
Pictures come from the media library. A visual question's picture must be one uploaded to the Next Exams media library — PNG, JPEG, GIF or WebP — and needs a description. A picture hosted somewhere else is refused.
Note: a puzzle shows the learner the pieces of a picture that their browser has to load whole, so the finished picture can be found by anyone who looks for it. Use puzzles as an engaging exercise, not as a secure item.
Results show each visual answer against the correct one: the pin and the areas, or the drawing and the region, drawn on the picture, with the same answer in words beside it.
How answers are stored
Next Exams uses a deliberate hybrid, because the two halves of a question have different lifetimes.
| What | Where it lives | Why |
|---|---|---|
| Choice options | Rows in their own table, each with a stable identifier | Attempt responses reference option identifiers directly. Stable rows mean a result stays meaningful even after an option's text is edited. |
| Grading configuration | A compact answer key stored with the question | Accepted answers, matching rules, tolerances and credit policy are not tabular, and do not need a row each. |
The practical benefit is that a new question type needs a grading case and an answer-key shape — not a schema change.
Grading rules per type
The answer key holds only what a given type actually needs.
| Setting | Applies to | What it does |
|---|---|---|
partial_credit |
Multiple choice (multiple), ordering, matching, EMQ, cloze | Awards a share of the points for a partly-correct answer instead of nothing. |
partial_policy |
The same types | all_or_nothing (the default) or proportional. |
accepted |
Short answer | The list of answers treated as correct. |
trim |
Short answer | Ignores leading and trailing whitespace. |
case_insensitive |
Short answer | Ignores letter case when comparing. |
use_regex |
Short answer | Treats each accepted answer as a regular expression rather than a literal. |
numeric |
Short answer | Compares as a number rather than a string. |
tolerance |
Short answer, numeric mode | Accepts anything within this margin of the expected value. |
Note: an empty answer and a wrong answer are not the same thing, and Next Exams is careful about the difference. The values
0,"0"andfalseare all real answers, not emptiness — a learner who answers zero has answered.
Subjects
Every question is filed against a subject. Subjects form a tree: each has an optional parent, and the component maintains each node's full path and depth automatically.
- Reorganising is safe. Moving a subject rebuilds the paths of everything beneath it in one operation, and a move that would make a subject its own ancestor is rejected.
- Subjects drive pools. A pool rule that draws from a subject is the most common way to build a randomised exam.
- Subjects drive feedback. When an exam spans two or more subjects, the result page breaks the score down per subject, weakest first, so the learner knows what to revise.
Tags
Tags are free-form labels, independent of the subject tree. Where a subject answers "what is this question about", a tag answers anything else you want to slice by — difficulty, syllabus reference, exam board, year.
Tags are useful for two things in particular:
- Filtering the bank while authoring, alongside subject, type and published state.
- Tag-driven pools — an exam can draw its random questions from a tag rather than a subject, which lets one pool cut across the subject tree.
Tags are a switchable app. Turning them off removes the authoring interface; existing tags remain in the database.
Custom fields
If the built-in fields do not capture something you need, add your own. Custom fields apply to two contexts — questions and exams — and are merged into the editing form at runtime, so they appear alongside the native fields rather than in a separate panel.
| Field behaviour | Where the value is stored |
|---|---|
| Marked searchable | Its own value table, so it can be filtered and searched efficiently |
| Not searchable | A JSON column on the question or exam itself |
There is deliberately no file field type — Next Exams does not accept arbitrary uploads through custom fields. Images belong in the media library, which validates them properly.
Rich text, maths and media
Question stems, options and explanations are authored in a rich text editor built on the Joomla installation's own bundled TinyMCE, so it inherits your editor configuration.
The editor is one page: the question and its answers on the left, its settings on the right. The Exam Studio uses the very same editor, so a question is written the same way wherever you write it.

Mathematics. LaTeX is typeset with KaTeX, in stems, options, explanations and on the result page. Maths is a switchable app, and the library is loaded only on pages whose content actually contains notation — a catalogue page with no formulae never downloads it.
Images. The editor's image button opens the Next Exams media picker rather than Joomla's general media manager. It offers a curated library, an upload, and an external URL, and:
- Uploads are raster only — PNG, JPEG, GIF and WebP. SVG is excluded deliberately, because an SVG can carry script.
- Every upload is validated by extension and by inspecting the file's actual image data, so a script renamed
.pngis rejected. - Alt text is required. The Insert button stays disabled until alt text is supplied.
The alt-text requirement is also enforced when the question is saved, not only in the picker, so an image pasted in as raw HTML cannot bypass it. Administrators can relax this for legacy content with a setting, but it is on by default.
Versioning and approval
Questions can carry a version history, with an approval workflow for changes.
- Each save records a version, so you can see how a question has evolved and who changed it. Version history in the question's header opens the list.
- Changes can require review before they go live, which matters when several authors share a bank.
- Attempts are unaffected either way: the paper is frozen when an examinee starts, so a question edited mid-exam never changes under them.
Versioning is a switchable app.
Questions that are in use
A question that exams use, or that learners have answered, is protected. The question list's Used in column shows how many exams use each one. Its page says so at the top — which exams use it, each a link to that exam, and whether learners have answered it — so you know before you save what the change will reach. (Someone without the right to view exams is told how many exams, not which.)
| Situation | What happens |
|---|---|
| Deleting a question an exam still uses | Refused, naming the exams. Take it out of them first. |
| Deleting a question learners have answered, that no exam uses any more | It is retired instead: hidden from the bank and from new exams, kept only so past results still open. |
| Rewording an answered question | Applied in place. |
| Changing the answers of an answered question | Saved as a new version that replaces it in the exams you choose. Past results keep the version that was answered. |
| Changing the type of an answered question | Refused. |
| Two people saving the same question | The second save is stopped; reload their version or overwrite it. |
Editing a question never disturbs answers already given: each option keeps its identity when the question is saved, so an answer recorded before an edit is still graded correctly after it.
Access control
| Action | Required permission |
|---|---|
| View the question bank | questions — view |
| Create a question | questions — create |
| Edit a question | questions — edit |
| Delete a question | questions — delete |
| Upload to the media library | questions — create |
| Manage subjects | subjects — the matching action |
Reads are gated as strictly as writes. A user without questions view permission cannot list questions, and cannot reach a single question's answer key by requesting it directly.