Virtual Reality Research Group

Distortions of visual space in an expanding virtual room

We have been investigating how people's perception of 3D space is affected when the scene around them expands or contracts (see our Current Biology paper). Remarkably, it turns out that people do not notice anything odd is happening. This has important implications for understanding how humans represent the 3D layout of a scene. Here, we describe some of the information that is normally available to people about the size and distance of objects and why much of that information might be ignored when they are in an expanding room.

Finding the distance to an object from multiple views
Expanding room experiment
The importance of a ground plane

We use immersive virtual reality to create environments that would be impossible to investigate otherwise, but which can tell us a lot about the signals the visual system uses to represent 3D space. People wear a head mounted display and walk around a virtual environment. In theory, this should give them lots of information about the 3D layout of the scene around them. Indeed, the sense of being in another world is compelling. However, no-one who has been in our expanding room has noticed anything odd happening when the room around them gets larger or smaller. We explain here some of the information that is normally available to people about the size and distance of objects as they walk around and why that information might be ignored in the expanding virtual room.

Finding the distance to an object from multiple views

A single image cannot tell you about the distance of objects. You might be looking at a flat picture, after all, rather than a real scene. But when you view a scene with two eyes (or move your head) you can tell whether or not you are looking at a flat picture. It is easy to understand why. A plan view and side view of a building tell you the 3D location of objects whereas a plan or side view on their own would not. Similarly, as the figure below shows, two views of an object can, in theory, tell you how far away it is.

Fig 1: Two eyes, drawn from above, view an object. To discover how far away it is (d), the visual system must know the direction of the object as seen by each eye and the separation (s) between the eyes.

To find the distance (d) of a single object, the visual system needs to know the visual direction of the object in each eye and the separation between the eyes (s). There are arguments about how good the visual system is at getting distance information in this way, especially when there is only one object in the scene as shown here, but there is plenty of evidence that the visual system uses an estimate of the distance between the eyes in order to judge the distance, size and shape of objects. One example was demonstrated by Helmholtz (1866) using a device that artificially increased the separation of the eyes. It had the effect of making everything appear small, "as if the observer were looking not at the natural landscape itself, but at a very exquisite and exact model of it, reduced in scale" implying that the visual system assumed the separation between the eyes, s, had not changed.

The same principles apply to moving your head. Here, you get many views of a scene. To work out how far away an object is from these images, the visual system still needs to know the visual direction of the object at each instant and the distance the eye has moved (s).

Fig 2: A 3D version of Fig. 1.

An expanding room

The expanding room experiment (see our Current Biology paper) exploits the unique attributes of virtual reality equipment, creating a virtual room that changes size. The centre of expansion is a point half way between the eyes. This is important since it means that the retinal projection of a virtual object remains the same irrespective of its size. For example, a wall of the room may double its distance from you but will also double in size. Alternatively, the moon is 1/400th the size of the sun, but appears to be the same size in the sky. This is because the sun is approximately 400 times further away.

As an observer walks around in the real room, the virtual room expands or contracts depending on how far they are from the left-hand wall. When they are in the middle of the real room, the real and virtual rooms are the same size so the observer's feet (which they cannot see) are at the same height as the virtual floor. On the left of the real room, the virtual room is half the size. On the right hand side of the real room, the virtual room has expanded to double its original size. Thus, walking from the extreme left to the extreme right of the physical room, the observer will see a four-fold expansion of the virtual room.

Fig 3: As the subject walks to to their left, the virtual room shrinks. As they walk to their right, it expands until it is four times the size. The red cubes are only visible part of the time and are used to test the observer's perception of size (see text).

Despite this massive change in scale, the pattern of light falling on the observer's retina is similar to that experienced by an observer walking through a static room, although the relationship between distance walked and image change is altered. Only stereopsis and motion parallax can give the observer any clue as to the size of the room they occupy. These are demonstrably powerful cues under normal circumstances and contribute to our ability to grasp different sized coffee mugs, navigate across furniture filled rooms without colliding with tables, and catch balls.

So, just how robust are stereopsis and motion parallax in an expanding room? We asked observers to make perform a simple size matching task. On the left hand size of the room (when it was half its normal size) they saw a red cube of a fixed size (5cm in the real room). We call this the reference cube. They then walked across to the right-hand side of the room, with the virtual room expanding such that it was now four times the size. As they walked, the reference cube disappeared, and a second test cube appeared. The size of the test cube varied from trial to trial, sometimes larger and sometimes smaller than the reference cube. The observer then had to say whether they thought the second (test) cube was bigger or smaller than the first (reference) cube. The computer logged their response and the next trial began, with the observer walking back to the left hand side of the room to observe the first cube again. Observer's would repeat this procedure several hundred times, each time with a different sized test cube, at one of three different distances. Eventually the computer 'homes in' on a test cube size that the observer perceives to be the same as the reference cube.

Fig 4: MPEG movie of the standard expanding room (click image to start movie). This looks just like a normal room, however if you look carefully the observer seems to move faster when they are in the left-hand side of the room than when they are on the right. In fact, they walked the same speed throughout. This is a consequence of the changing size of the virtual room.

We found that observers made gross errors in size judgements, sometimes indicating that a cube four times larger than the reference was the same size. We tested this phenomena at different distances (placing the test cube at one of three different distances from the observer). While observers were better at judging the size of nearer cubes, they still made judgement errors by a factor of two.

This is a profound result. Stereopsis and motion parallax are known to be two of the most powerful cues to 3D vision and yet we are largely ignoring them. Questions it raises include:

Why does the brain place so much weight on the assumption the scene remains a constant size?
Why is the brain so willing to throw away correct information from other visual and non-visual cues?
Can we kick the brain into using the correct cues and ignore the misleading ones?

Our research has turned to answering these questions. For example, what if we placed non-expanding objects in the virtual room: objects that remained the same size in the real world, irrespective of the size of the expanding room?

In this example, we placed humanoid figures into the expanding room and asked observer's to repeat the experiment. The figures did not move, and were placed on large pedestals such that they appeared to be `placed' on the floor in the large expanding room rather than just floating in space.

Fig 5: MPEG movie of the expanding room with constant sized human figures (click image to start movie).

Observers were much better at matching sizes of objects in this condition. In fact, their improvement carried over the original expanding room shown in Fig. 4.

The movie below shows how much the room expanded. In this variation of the experiment, the bricks remain the same size in the real world, with extra bricks `sliding' into view as the room walls increase in size. Now it seems to the observer that they are in a small brick closet on the left-hand side of the room, which grows into a large sports hall on the right-hand side. In this case, of course, people are very good at judging the size of objects. Unlike the experiment with the human figures, however, their improvement did not transfer to the original expanding room condition (see Fig. 4).

Fig 6: MPEG movie of the expanding Room with constant sized bricks (click image to start movie).

The importance of a ground plane

The floor can be a very important cure to signal how far away objects are, because normally our eyes are a constant height above the ground. We have done various experiments with the floor removed to examine, for example, how powerful stereopsis and motion parallax can be when they do not conflict with other assumptions or cues. A movie from one such experiments is shown is Fig 7, below.

Fig 7: MPEG movie (click image to start movie) of an experiment in which the floor has been removed. Observers moved from a small static room to a large static room and made the same size judgement as before. The textures on the wall cannot help the observer to judge the relative sizes of the rooms.

Page last updated on 11:06:08.