The reality–virtuality continuum is a conceptual scale that places every display environment between two poles: the unmodified real environment at one end and a fully computer-generated virtual environment at the other. Among the XR fundamentals of extended reality, it is the reference model that explains why augmented reality, mixed reality and virtual reality are points on one spectrum rather than unrelated technologies.
The model was introduced by Paul Milgram and Fumio Kishino in 1994 as part of a taxonomy of mixed reality visual displays. Its central claim is that the meaningful variable is how much of what the user perceives is modeled by the computer and how much comes directly from the physical world.
How the continuum orders real and virtual content
The continuum orders environments by the proportion and dominance of synthetic content in the user’s perception. Everything between the two endpoints is what Milgram and Kishino called mixed reality, a broad region in which real and virtual elements are presented together in one display.
Inside that region, the original paper distinguished two sub-zones. In augmented reality, the primary experience is the real world and virtual objects are added to it. In augmented virtuality, the primary experience is a virtual world into which real elements, such as a live video of the user’s hands or a scanned room, are inserted. The boundary between the two is deliberately blurry, because the scale is continuous.
The following table summarizes the main positions on the scale and what dominates the user’s view at each one.
| Position | Dominant content | Typical example |
|---|---|---|
| Real environment | Physical world only | Looking through a clear window |
| Augmented reality | Real world with virtual overlays | Navigation arrows projected on a street view |
| Augmented virtuality | Virtual world with real inserts | Virtual cockpit showing the pilot’s real hands |
| Virtual environment | Synthetic world only | Fully rendered VR game level |
The three dimensions behind the original taxonomy
The continuum is only the most cited part of a richer framework. Milgram and Kishino argued that a single axis could not describe a display completely, so they proposed three supporting dimensions that explain where on the scale a system can realistically sit.
Extent of world knowledge describes how much the system knows about the objects and geometry of the real scene. Reproduction fidelity describes how realistically virtual content is rendered. Extent of presence metaphor describes how strongly the interface makes the user feel located inside the displayed scene, from a flat monitor to a fully immersive head-mounted display.
| Dimension | Low end | High end | Why it matters |
|---|---|---|---|
| Extent of world knowledge | Unmodeled scene | Every object modeled | Determines occlusion and anchoring accuracy |
| Reproduction fidelity | Wireframe graphics | Photorealistic rendering | Affects how convincingly virtual objects blend |
| Extent of presence metaphor | Monitor-based window | Real-time immersive view | Shapes immersion and sense of presence |
These dimensions explain a practical point: placing virtual furniture convincingly in a real room requires high world knowledge, because the system must understand floors, walls and lighting. Modern spatial mapping and scene understanding in headsets are, in effect, engineering answers to that dimension.
How AR, MR, VR and XR map onto the scale
Current industry vocabulary maps onto the continuum, but not always in the way the original authors intended. In the 1994 paper, mixed reality meant the entire middle region. Many vendors now use MR to mean a specific class of experiences in which virtual objects are spatially anchored and interact with real surfaces, which narrows the original meaning.
Extended reality (XR) is an umbrella label covering the whole continuum, including its virtual endpoint. Spatial computing describes a computing paradigm rather than a position on the scale, although it usually relies on mixed reality capabilities.
| Term | Coverage on the continuum | Typical display approach |
|---|---|---|
| AR | Near the real end | Optical see-through glasses or phone camera view |
| MR | Middle region with anchored content | Video see-through or optical see-through headset |
| Augmented virtuality | Near the virtual end | VR headset with real-world inserts |
| VR | Virtual endpoint | Opaque head-mounted display |
| XR | Entire continuum | Any of the above |
Display technology influences position but does not fix it. An optical see-through device shows the real world directly through transparent optics, while a video see-through device captures the world with cameras and redisplays it. A video see-through headset can therefore move freely along the scale, from passthrough-heavy AR to full VR, simply by changing how much camera imagery it shows.
Using the continuum to design and evaluate experiences
Designers use the continuum as a planning tool for choosing the right blend of real and virtual content for a task. The key question is whether the user needs continuous awareness of the physical environment or benefits from full isolation from it.
| Situation | Recommended position | Reason |
|---|---|---|
| Maintenance on real machinery | Augmented reality | Technician must see and touch the real equipment |
| Collaborative design review in an office | Mixed reality | Virtual model shares space with colleagues and tables |
| Flight simulator with physical controls | Augmented virtuality | Virtual scenery with real hands and switches visible |
| Exposure therapy or immersive training | Virtual reality | Full control over every stimulus |
The model also helps evaluate devices. A headset marketed as mixed reality can be judged by how far along the scale it can travel and how well it handles the hard middle region, where errors in tracking, latency or lighting estimation become most visible.
Limitations and common misconceptions
The continuum is a conceptual model, not a measurement scale. It has no units, and two experts can place the same application at different points. It is best read as an ordering principle rather than a precise metric.
- It is visual by origin. The 1994 framework focused on displays; audio, haptics and smell can also be mixed, and later researchers proposed multisensory extensions.
- The endpoints are idealized. Even the most immersive VR still involves the real body, real gravity and real floor contact, so a truly pure virtual environment does not exist in practice.
- MR is not a separate device category by definition. Treating MR as a product class obscures the original idea that it spans most of the scale.
- Position is not quality. A point closer to the virtual end is not more advanced; it simply serves different tasks.
Some researchers also criticize the model for ignoring interaction and user agency, since it describes what is displayed rather than what users can do. Frameworks for presence and immersion are therefore used alongside it, not replaced by it.
In simple terms
Picture a slider. At one end you see the world exactly as it is. At the other end everything you see is made by a computer. The reality–virtuality continuum is that slider.
Phone apps that put a digital sticker on your table sit close to the real end. A headset game that replaces the whole room sits at the virtual end. Everything in between mixes the two. The key idea is that AR, MR and VR are not rivals but different settings on the same slider.
A practical example
An aircraft maintenance training center plans a new program for engine inspection. The team uses the continuum to decide where each module belongs. For the first lessons, trainees wear a fully immersive VR headset and explore a virtual engine, because mistakes cost nothing and instructors can trigger rare faults on demand.
Later modules shift toward the real end. Trainees stand next to a real engine and wear a video see-through headset that overlays torque values, part labels and step-by-step arrows onto the physical components. The final assessment uses no overlays at all.
By moving students deliberately along the scale, the center shortens time to hands-on readiness while keeping access to the real engine, its most expensive resource, for the stages where it matters most.