THE GOLD STANDARD · VIRTUAL REALITY
Understand the world inside the headset
Virtual reality is useful when its representation, interaction and learning purpose fit together. This guide explains the technology from first principles, then shows how to test what a learner can understand and use afterwards. The worked examples need no headset.
Find your next learning step
Choose one route or expand a contents group. Every teaching chapter remains open below the index.
Read the full guide · Practice and worked answers · Education learning hub
Understand the technology · Chapters 1–5
Design for people and learning · Chapters 6–10
Evaluate, practise and connect · Chapters 11–15
CHAPTER 1 OF 15 · Understand the system
1. What virtual reality is—and what it is not
Virtual reality, or VR, presents a computer-mediated environment that can respond to a participant’s actions and viewpoint. In an immersive headset experience, turning the head changes the view into the represented space. Some systems also track position, controllers or hands. The defining question is not whether the picture is impressive, but how the environment is presented and how the person can interact with it.
Augmented reality adds digital content to a view of the physical environment. Mixed reality is used for experiences combining physical and digital elements, although product terminology varies. Extended reality, or XR, is a useful umbrella term. Do not assume these labels specify the same tracking, display, privacy or access requirements. Describe what the particular system actually does.
A panoramic video can let someone look around from a recorded viewpoint without letting them walk freely through the recorded scene. A desktop three-dimensional model can permit rich manipulation without surrounding the viewer. Both may be educationally useful. Calling everything VR conceals important differences between looking around, changing position and changing the model itself.
This root connects the technology to learning, representation and responsible design. It is not a buying guide, a device manual or permission to use a headset with a particular child. Equipment requirements, institutional safeguards and individual needs must be checked separately. Its worked examples use invented numbers to make the reasoning inspectable.
The Khronos OpenXR overview introduces a cross-platform interface for XR applications and devices. A common interface can reduce some platform-specific work; it does not make every device, feature or experience interchangeable.
CHAPTER 2 OF 15 · Understand the system
2. Follow the loop from movement to a displayed world
Imagine a learner examining a virtual sculpture. They turn their head to the right. The system needs an estimate of the new viewpoint, an updated representation of the scene and an appropriate displayed image. If the sculpture is meant to stay in one place, its apparent position must change consistently with that viewpoint. Otherwise it can seem attached to the person’s face rather than located in the room.
A simplified loop is: estimate the relevant pose, update the application, prepare the views, render images and present the result. A pose combines position and orientation. This account is deliberately simplified: actual implementations can predict display-time poses and perform additional processing. The useful first principle is that movement and presentation must remain coherently related.
Consider three faults separately. The sculpture may jump because tracking becomes unreliable. A button may fail because the application does not recognise an input. The whole scene may update unevenly because the rendering workload is excessive. They can all feel like a bad experience, but they require different investigations.
Write observations before diagnoses. ‘The sculpture moved when the participant leaned sideways’ is an observation. ‘The tracker is broken’ is an interpretation that needs testing. Repeat the action, inspect the relevant system information and compare with another scene or input. Do not change several unrelated settings and then claim to know which one repaired the problem.
This system view connects to Human–Computer Interaction: the participant needs understandable actions, useful feedback and a dependable route out when something goes wrong.
CHAPTER 3 OF 15 · Understand the system
3. Rotation, translation and reference spaces
Looking left and stepping left are different movements. Rotation changes orientation. Translation changes position. A system that responds to head rotation alone cannot provide the same viewpoint changes as a system that also tracks translation. This distinction explains why some experiences support looking around from one location while others support leaning and moving around objects.
A coordinate is meaningful only with its reference frame and units. The same sculpture can have coordinates relative to the room, a movable platform or the viewer. Mixing those frames is like describing a seat’s position relative to a bus and then interpreting the numbers as its position along the road. The numbers may be valid while the interpretation is wrong.
Use a one-dimensional teaching example. A virtual marker is at x = 2.0 metres in a room frame. The viewer is at x = 0.5 metres. With axes aligned and no rotation, the marker is 2.0 − 0.5 = 1.5 metres to the viewer’s positive-x side. If the viewer moves to x = 0.8 metres while the marker stays fixed, that relative distance becomes 1.2 metres.
This subtraction is not a complete three-dimensional camera transformation. Rotated frames require orientation to be accounted for as well. Nor does a stable-looking coordinate prove that physical tracking is exact. The OpenXR specification’s spaces chapter explains reference spaces and why their relationships can be adjusted as tracking evolves.
The learning checkpoint is simple: state what moves, what stays fixed and which frame each quantity belongs to. Continue to Vectors for quantities with direction. A correct numerical answer without a reference frame can still describe the wrong relationship.
Changing a reference origin requires more than reusing the earlier coordinate numbers. Keeping an Anchored Object Fixed When a Reference Space Changes derives original position-and-orientation updates, proves the viewer-relative pose invariant and checks transform order, time, validity and independent changed cases.
CHAPTER 4 OF 15 · Inspect the representation
4. How a three-dimensional world becomes a view
A virtual object is a representation, not a physical object hidden inside the headset. Its geometry, surface appearance, lighting and viewpoint contribute to the image. Different views for the two eyes can provide stereoscopic information, while perspective, occlusion and changes during movement can also convey depth. A convincing view does not by itself establish accurate scale or scientific truth.
Suppose an exhibition model contains a sphere labelled as a planet. The designer enlarges it so visitors can examine its surface. That can be a useful teaching decision, provided the enlargement is explicit. If nearby objects remain at another scale, distances in the scene cannot be read as literal astronomical distances.
Distinguish visual fidelity from explanatory fidelity. A highly detailed machine may conceal the component whose operation matters. A simplified transparent model may explain that operation more clearly. Conversely, removing a connection essential to the mechanism makes the explanation wrong even if the simplified display looks tidy.
For a fictional gallery, ask three separate questions. Does the model resemble the sculpture? Are its dimensions suitable for judging proportions? Is the surrounding arrangement historically documented or reconstructed? A viewer should be able to distinguish a scan, an interpretation and a deliberately altered teaching model.
Optics develops the behaviour of light and image formation. Visual Communication develops the choices that make a representation understandable. Neither realistic lighting nor beautiful graphics excuses an unexplained change of scale.
CHAPTER 5 OF 15 · Inspect the representation
5. Worked example: frame time is a budget
A display updating at ninety refreshes per second has an interval of 1 ÷ 90 seconds, approximately 11.11 milliseconds. At seventy-two refreshes per second the interval is approximately 13.89 milliseconds. These are arithmetic examples, not universal target rates or claims about a particular device.
Now suppose a simplified measurement reports a graphics workload of nine milliseconds per frame. Relative to an 11.11-millisecond interval, the nominal difference is about 2.11 milliseconds. That is not a guarantee of smooth operation: other work, scheduling, runtime processing and variation still matter. Treat it as a first comparison, not a complete latency measurement.
If a heavier scene takes thirteen milliseconds under the same simplified conditions, it exceeds that interval by about 1.89 milliseconds. Improving average performance alone may not remove occasional slow frames. Record behaviour in demanding scenes and during actual interactions rather than measuring only an empty starting room.
Be careful when adding timing figures. CPU and GPU work may overlap; their reported durations are not necessarily consecutive blocks that can simply be summed. The Unity frame-timing documentation explains engine-specific timing and profiling considerations. Its details should be read for the relevant software version.
Google’s Android XR frame-budget guidance likewise treats rendering time as a practical constraint. The transferable method is to define the target, measure on the intended system, identify the bottleneck and test a controlled change. A frame interval is not the same thing as the complete delay between physical movement and visible response.
| Illustrative update rate | Interval calculation | Approximate interval |
|---|---|---|
| 72 per second | 1,000 ÷ 72 | 13.89 ms |
| 80 per second | 1,000 ÷ 80 | 12.50 ms |
| 90 per second | 1,000 ÷ 90 | 11.11 ms |
Separating Pose Sample Age from Rendering Time follows a fictional common-clock frame from source capture to defined presentation, showing why overlapping render work, prediction targets and the age of pose information answer different questions.
CHAPTER 6 OF 15 · Inspect the representation
6. Interaction should support the task, not become the task
An immersive interface can use controllers, hands, gaze or other inputs, depending on the system. The best choice cannot be inferred from novelty alone. Identify the action the participant needs: select a sculpture, rotate a model, compare two views or record an observation. Then ask whether the input makes that action understandable and recoverable.
In the fictional gallery, a learner reaches towards a label but accidentally moves the sculpture. The failure is not simply that the learner needs more confidence. Selection and manipulation may be insufficiently distinguished. A design change might separate the label action from object movement and provide immediate feedback about which mode is active.
Test cancellation and recovery as seriously as successful selection. Can the participant undo a rotation, return an object to its original orientation and reopen instructions? Can they leave an activity without losing work unexpectedly? A demonstration that follows one perfect sequence does not reveal what happens when a person makes a reasonable mistake.
Instructions should describe the available action in language consistent with the actual input. ‘Press the left button’ becomes ambiguous if controllers are remapped or a participant uses an alternative input. Action-based wording and accessible guidance are more adaptable than assuming everyone holds the same hardware in the same way.
Information organisation matters inside a spatial interface too. Persistent instructions, optional detail and task feedback have different jobs. Information Architecture helps clarify where each belongs so that the participant does not have to search a whole virtual room for a basic instruction.
CHAPTER 7 OF 15 · Plan participation and learning
7. Choose movement and access deliberately
A virtual destination may be reached by walking physically, selecting a teleport destination or using an artificial movement control. These are not equivalent experiences. Each changes the relationship between the person’s physical location, virtual viewpoint and required actions. A system should explain its movement model before asking someone to navigate a demanding task.
The physical room remains real even when it is not visible through the displayed scene. A rendered clear path is not evidence that the corresponding physical area is clear. Actual equipment instructions, supervision and the authorised environment govern use. Virtual boundaries or passthrough features should not be treated as substitutes for the full applicable safeguards.
Design the learning task so that pausing or stopping is straightforward and does not carry an academic penalty. Do not require someone to continue an uncomfortable experience to finish a graded task. A non-headset route should be planned before the session rather than improvised after a participant cannot continue.
Accessibility is not one setting. A task might depend on small text, sound direction, sustained standing, two-handed input or reaching above shoulder height. Name the dependency, then determine how the same meaningful action can be supported in another way. A seated mode alone does not solve an audio-only instruction or an inaccessible menu.
The W3C XR Accessibility User Requirements discusses needs including alternative interaction, personalisation and multimodal information. It is a Working Group Note, not a certificate that a product meets every user’s needs. Universal Design extends the question from individual controls to the complete participation experience.
CHAPTER 8 OF 15 · Plan participation and learning
8. Presence is not proof of learning
A participant can feel located inside a represented environment while learning little about its subject. They may remember the size of a virtual room but not the relationship the lesson intended to explain. Enjoyment, presence, usability and learning are different outcomes. Measure them separately instead of using one as a convenient substitute for another.
A 2022 study by Ahn, Nowak and Bailenson reported that heightened spatial presence could accompany reduced recall in the experiments studied. This does not establish that VR generally harms learning. It demonstrates why a stronger sense of being there cannot automatically be interpreted as stronger understanding.
Begin with a precise learning outcome. ‘Explore the gallery’ describes an activity. ‘Explain how the same sculpture’s apparent outline changes with viewpoint while its shape stays fixed’ describes understanding that can be checked. The latter makes it possible to compare a headset experience with photographs, a physical model or a desktop viewer.
Teach the required vocabulary and basic controls before combining them with the main reasoning task. If a learner is still trying to find the selection control, their difficulty answering a subject question is hard to interpret. Separate an interface rehearsal from an assessment of the concept.
Afterwards, remove the familiar visual prompts. Ask the learner to explain a new example, sketch a relationship or identify a misleading representation. A correct answer inside a scene that highlights the right object may reflect successful cue-following rather than independent understanding. The gold standard is what remains usable when those cues change.
CHAPTER 9 OF 15 · Plan participation and learning
9. Design an original lesson around a real explanatory need
Consider a hypothetical lesson about viewpoint and shape. Its objective is for learners to distinguish an object’s three-dimensional form from one two-dimensional outline. The lesson could use a simple asymmetrical sculpture, not an ornate environment whose details compete with that objective.
Before the activity, show a still image and ask learners what the hidden side might look like. Record explanations rather than treating guesses as failures. During the activity, ask them to predict how the outline will change before moving to a second viewpoint. Then compare the prediction with the observed view and identify what was missing from the original image.
The teacher’s explanation should connect the actions to the idea: the viewpoint changed, the object did not deform, and one projection does not uniquely reveal every surface. A control that accidentally changes the object’s shape would confound this lesson unless that change were explicitly part of a later comparison.
After the activity, present a different object using two ordinary drawings. Ask which features can be inferred confidently and which remain uncertain. This is a transfer task because the learner must use the concept outside the original immersive setting. A physical model or accessible desktop interaction can support the same reasoning where appropriate.
Specify the fallback in advance. If the headset fails, a sequence of verified views and a manipulable physical object can preserve the comparison. The lesson’s purpose should survive a hardware failure. For a broader classroom assurance route, read Classroom AR and VR: Learning, Device Safety and Accessibility; it is an independent eduKate guide, not an official ministry instruction.
CHAPTER 10 OF 15 · Test limits and evidence
10. Simulation fidelity, training and the limits of transfer
A simulation represents selected aspects of a situation. It cannot include every condition of the world it represents. The useful question is which relationships must be preserved for the intended task. Visual resemblance is one kind of fidelity; realistic timing, forces, information and consequences can matter differently.
Suppose a virtual assembly task teaches the order in which labelled pieces fit together. It might support learning that sequence even if the controls use a button instead of a physical tool. It would not, on that basis alone, establish competence in applying the correct force with the real tool. The missing demand is central to the second claim.
Distinguish practice success from real-world authorisation. Completing a simulated procedure does not by itself qualify a person to operate hazardous equipment or perform clinical work. Such use requires the relevant professional standards, validated training and authorised assessment. A teaching demonstration should not imply otherwise.
Transfer testing should change a relevant surface feature while preserving the underlying principle. Use a different object arrangement, remove a label or introduce a new case. Then check whether the person explains what must stay the same and what must change. Repeating the identical virtual sequence can measure memorisation of that sequence.
Also inspect negative transfer: a convenient shortcut inside the simulation may be inappropriate outside it. If virtual objects pass through one another or reset without consequences, make those conventions explicit. An effective debrief names both what the model teaches and what it deliberately leaves out.
CHAPTER 11 OF 15 · Test limits and evidence
11. Privacy and shared spaces belong in the design
Immersive systems may process information beyond typed answers, including pose information needed to present the experience. Other capabilities can include voice, cameras, room information or eye tracking, depending on the device and application. Do not assume every system collects all of these, or that all processing means permanent remote storage.
The W3C WebXR Device API specification discusses sensitive pose information and security, privacy and comfort considerations. It helps explain why spatial tracking is not merely an ordinary page preference. The document is a Candidate Recommendation Draft as reviewed for this article; implementation and platform support must be checked separately.
An educational review should identify what information is necessary, what is retained, who can access it and what happens when the session ends. A permission prompt is only one part of that review. Ask whether a recording is needed at all rather than treating every available data field as useful evidence.
In a shared session, participants also need understandable identity, communication and exit arrangements. A familiar avatar name is not proof of identity. A public room is not equivalent to a controlled class group. Specify who can enter, who can speak or record and how a participant gets help without making their private circumstances public.
Keep account access and device management within approved arrangements. Learners should not need to disclose unrelated personal information to complete a lesson. Cybersecurity provides the wider framework for protecting systems and accounts; privacy also requires a justified reason to collect and use information in the first place.
CHAPTER 12 OF 15 · Test limits and evidence
12. Evaluate a pilot without overselling the result
A pilot should answer a defined question: can participants complete the target task, and does the experience support the intended learning? Record interface difficulties separately from subject errors. Note who could not participate or needed an alternative. Excluding those people from the report can make an apparently successful experience look more inclusive than it was.
Use an invented example. Twelve learners take a concept test before an activity, averaging four out of ten, and afterwards average seven out of ten. The observed mean increase is three points. It is not evidence that VR alone caused a three-point gain. Practice with the test, teaching, prior differences and other conditions could contribute.
Now suppose another group using a desktop model increases from five to seven. Its observed increase is two points. The difference between group changes is one point, but this subtraction still does not guarantee a causal effect. Assignment, comparable instruction, assessment quality and uncertainty matter. Small groups can also produce unstable averages.
Include a delayed or changed-context assessment if durable transfer is the aim. Immediate recall of highlighted labels answers a narrower question. Collect qualitative observations to explain patterns: several people may have missed a concept because a menu blocked the view, while others may have understood it through the non-headset route.
Make the next decision proportional to the evidence. A pilot can justify repairing an interaction and testing again without justifying a claim that the whole school should replace another teaching method. Statistics helps separate a description of this group from a confident claim about a wider population.
| Hypothetical group | Before /10 | After /10 | Mean change |
|---|---|---|---|
| VR activity | 4 | 7 | +3 points |
| Desktop activity | 5 | 7 | +2 points |
CHAPTER 13 OF 15 · Practise and connect
13. Practise the reasoning and repair the first unstable step
Practice A: a hypothetical display has an update interval based on eighty refreshes per second. Calculate the interval in milliseconds. Compare it with a simplified eleven-millisecond graphics workload. State one reason that the remaining interval is not a guarantee of low end-to-end delay.
Practice B: in an aligned one-dimensional room frame, a marker is at x = 3.0 metres and a viewer is at x = 1.2 metres. Find the marker’s relative position. Then move the viewer to x = 1.7 metres without moving the marker. Explain whether the marker itself changed position in the room.
Practice C: a class reports that learners enjoyed a virtual gallery and completed every highlighted prompt. Design one task that would test understanding without the original highlights. Identify an accessible alternative that preserves the question being assessed rather than substituting an unrelated worksheet.
If a learner subtracts the coordinates in the wrong order, repair the meaning before adding more dimensions. Ask ‘position of which object, relative to which origin?’ and draw the simple line. If they confuse milliseconds and seconds, convert one second into one thousand milliseconds before returning to refresh rates.
Stabilise the method with a nearby variation, then extend it. Change the viewer’s position, use another update rate or remove a visual cue. Do not mistake one corrected answer for a secure explanation. The learner should be able to state the relationship and anticipate how the answer changes.
The Secondary 1 Mathematics learning guide provides a connected route into foundational reasoning. Here the same principle applies to technology: identify the exact missing idea, teach it clearly, check it and then use it in a changed situation.
Worked answers and checking
A: 1,000 ÷ 80 = 12.5 milliseconds. Subtracting an eleven-millisecond workload leaves a nominal 1.5 milliseconds. Scheduling, other work and frame-to-frame variation can still matter. This subtraction does not measure the whole movement-to-display pathway.
B: Initially, 3.0 − 1.2 = 1.8 metres in the positive-x direction relative to the viewer. Afterwards, 3.0 − 1.7 = 1.3 metres. The marker stayed at x = 3.0 metres in the room; its position relative to the viewer changed because the viewer moved. These results assume aligned axes and no rotation.
C: Show two ordinary views of a new object and ask the learner to explain which apparent changes come from viewpoint and which would require a change in shape. Remove the original highlights and ask for a reason. A manipulable physical object or accessible desktop viewer can preserve the same comparison. Record whether any alternative changes the demands of the assessment.
CHAPTER 14 OF 15 · Practise and connect
14. Questions worth asking before the next immersive session
Is a panoramic tour the same as moving through a three-dimensional model? No. Looking around a recorded viewpoint does not necessarily let the viewer translate to another viewpoint. Check the particular experience rather than relying on its label.
Does a higher refresh rate prove a better lesson? No. Timing is one technical property. Accurate content, understandable interaction, access and an appropriate learning task remain separate requirements. A technically smooth scene can teach the wrong idea.
Does OpenXR guarantee that an application works identically everywhere? No. A common interface helps interoperability, but supported capabilities, implementation and application testing still matter. Verify the actual device and software combination.
Can a more realistic simulation always improve transfer? No. Preserve the relationships relevant to the task and test transfer directly. Extra detail can be irrelevant, while a missing physical demand can be essential.
Should a learner have to use a headset to receive full credit? The assessment should target the intended knowledge or skill, with appropriate access arrangements. Difficulty using a particular input is not automatically a lack of subject understanding. Plan a meaningful alternative.
Can school staff infer a health condition from difficulty in VR? This guide does not support diagnosis or treatment. Follow the institution’s authorised procedures and relevant professional advice. Record observations without attaching unsupported clinical labels.
What is the first question for a proposed VR lesson? Ask what the learner should explain or do afterwards that is difficult to teach another way. If that answer is unclear, develop the learning task before choosing more elaborate technology.
CHAPTER 15 OF 15 · Practise and connect
15. Continue through the ecosystem and beyond the headset
This root belongs to The Gold Standard of Education: Learning Transfers into Life. Its central standard is a coherent relationship between representation, action and understanding. Immersion is a possible means of teaching, not an outcome that replaces learning.
Follow the classroom assurance guide for lesson operations and the HCI root for understandable interaction. Use Optics and Vectors for the physical and mathematical relationships, Visual Communication for explanatory choices, and Universal Design for participation. These routes develop different parts of one experience without pretending that one discipline can answer every question.
For cultural exploration, Google Arts & Culture for Students provides a connected route into virtual museums and source checking. A reconstructed room, digitised object and historical explanation should be evaluated on their evidence, not accepted simply because they are presented inside an engaging environment.
Sources were reviewed on 21 October 2026. Standards documents describe interfaces or user needs, engine documentation explains implementation details, and the cited research examines particular learning conditions. None is a universal approval for a device, age group or classroom arrangement. The numerical exercises are original hypothetical examples, not product benchmarks.
For a final transfer task, choose a simple model without using a headset. Describe its reference frame, one action a user can take and one thing the model leaves out. Explain how you would check the resulting understanding in a different setting. If that explanation is clear, you are evaluating virtual reality as a purposeful learning system rather than mistaking the display for the education.