This report is also available as an Acrobat file.

Contents
The Design of Virtual Environments with particular reference to VRML

Representation

The truth in photography

The flat, usually rectangular, images which are photographs make a claim to be true that painting can never make.
Sontag 1977 p86

Photographic images have a reputation for showing things as they really are. This is not the place to debate the truth or otherwise of photography, nor to dwell on the practices which can assist photography in being untruthful (Sontag p115 passim is useful on the various claims made for photography's relationship to knowledge). Many have stood in front of a painting of a historical figure and thought, 'I wonder what s/he really looked like?' and have had little doubt that a photograph would have helped to answer this question. Whatever the truth or otherwise of photography, when it comes to belief, it is unrivalled. Academics would encourage a student to treat with some scepticism a photograph from, say, an electron microscope, or from a historical archive of 1930s Britain, but not the same kind of scepticism that we would adopt towards a drawing. We believe in what photography shows us.

It is interesting to speculate what the 'belief value' of VR will be in future. While VRs remain clearly constructions, they are likely to be regarded as low in believability, however realistic (in another sense) they may look. But if a practical means can be made available whereby VR models can be created as a counterpart to reality (whether using multiple cameras, laser-scanning or whatever), in the same way that photographs are perceived as an unmediated response to the light-values of a scene, then the perceived authenticity of such models will receive a major fillip.

Of course the question of authorship and the status of the author will also be influential: the triad - authorship, authority, authenticity - will continue to be an issue as it is in relation to any medium.

Interestingly, the belief we have in photography seems to illicitly carry over into our appreciation of overtly fictional works, so that films are seen as in quite a significant way 'more real' than plays (despite the fact that in a play the actors are really there).

The truth in drawing

By 'drawing' here we mean any non-photographic image construction. Perspective drawing allows us to create 'convincing' representations of the future, and of things which are unseeable (such as atomic structures). This is one of the aspects of VR - which in this sense we can regard as a form of perspective drawing - which has excited most interest. Even when we forgo the realism about things seen which photography seems to offer, the feeling of truth to appearances which 3-point perspective drawings provide is one of their greatest advantages.

Selection in drawing and photography

Another important aspect of the relationship between the model stored and the scene rendered is that of selection. In photography or film/video, selection is achieved by a number of means, but especially by framing - deciding at what to point the camera. This is an authorial prerogative which we have deliberately denied ourselves in VR, where we normally allow users to look where they please. However, the issue of selection still arises. though in a different form.

Some VR systems may resemble photography in the respect that whatever appears in the environment (the stored model) will be seen (rendered). If the mapping of a complex 3-space construction onto the visual display is done wholesale, without any ability to suppress detail, to render some objects translucently or omit them altogether, then the user may be oppressed by irrelevant data. To avoid this, we will again want to take liberties with our virtual realities, deciding what to show and what to suppress, in essence, 'filtering' the model. Levels of detail go a small way toward implementing this concept.

One of the reasons why drawing and other forms of 'constructed' representation still have a role since photography is that they allow constructive filtering. The technical illustrator knows when to omit whole parts of a machine in order to reveal concealed mechanisms. The structural engineer strips away all the irrelevant aspects of a building to show the bare bones of a structure. Here we are touching again on the issue of abstraction - the omitting of inessentials and the representation only of what is important for the purpose - which is (or should be) of overriding importance in discussing the application of any technology to Education. It suggests that the effect of VR (and particular kinds of Virtual Environment and Virtual Object) on educational processes and learner perceptions requires more research.

Dynamics in drawing

We have been using the term 'drawing' for any constructed 3-point perspective representation which is not automated like photography. Turning to a more limited traditional view of 'drawing' - manual mark-making - there are some special merits that this kind of image has over both photography and computer-rendered models. In many cases, looking at a drawing allows us not just to see the final result, but to decipher the process which produced it.

Productive ambiguities in drawing

We are used to the idea that ambiguity in information is a problem, so the idea of productive ambiguity may be unfamiliar. However, if we consider drawing as a process, rather than a product, it becomes clear that there are merits to imprecision about what is depicted. Arnheim, in a useful summary (1993) offers some observations on the superiority of sketching over more concretised imagery. These include:

Sketches are 'guiding images' whose role as externalisation facilitates the design process rather than being an end in itself
'As long as the guiding image is still developing it remains tentative, generic, vague'
'This vagueness however is by no means a negative quality.' The sketch 'stands for a whole range of possibilities without being tangibly committed to any one of them.'

The relevance to VR is that, like most computer representations, Virtual Environments and Virtual Objects have a precision and a finality which may be difficult to resist - the 'real presence' which is one of their strengths could, used inappropriately, be their weakness.

Distortion

Distortion is a loaded word. It implies falsehood. We think we know the difference between a visually distorted and an undistorted image. However, 'distortion' can be a more subtle affair, which assists rather than impedes the user's experience. As so often, the history of painting provides insights, and Dubery and Willats (1972) give a number of useful examples. We cite two:

If we survey a wide angle on a scene, using 'undistorted' 3P-perspective, the resulting image tends to have what we subjectively regard as distortions at the edges - things seem overlarge, or stretched, at the perimeter of the view (Dubery and Willats 1972, p87-89). However, if we distort the objects, as though they were projected onto a curved picture-plane, we create what appears to be a more 'natural' view (ibid p89-90). In addition, we gain the advantage, as Van Gogh does in his painting Vincent's Room, that viewers 'sense' their own location in relation to the scene - a key requirement of virtuality. Vincent's Room feels like a small, intimate space, because of, not in spite of, the 'distortions' imposed on it.
The work of Cezanne offers another interesting challenge when we think of the relationship between the photographically correct and the 'feeling real', where accurate spatiality and convincing virtuality seem to make competing demands. Notoriously, Cezanne painted what he saw, in a different sense to that normally meant. Rather than make a painting which is a time-less summation of the whole scene as though apprehended in an instant, Cezanne attended to the process of seeing. Whereas a table-edge which is bisected by a jug is conventionally depicted as the straight line which we 'know' it to be, Cezanne would rather show the angle that the table-edge seems to make when observed on one side of the jug, and then the angle at the other side, even though, on looking overall, we can see that the two angles are inconsistent - that the lines if extrapolated do not correctly 'join up'.

We are not suggesting that either of these specific techniques should be applied in rendering virtual models to the display, but they are intriguing examples of visual distortion which serve the purpose of making things feel more real.

A more prosaic example is given in Sarkar and Brown (1994), where productive use is made of the distortions of 'fisheye' views. Within the restrictive space offered by a conventional display, their aim was to reconcile the demands of providing maximum contextual information, with the largest, clearest possible view of the current topic of interest. The material to be displayed was a network diagram of linked notes. The solution adopted was to provide a central zone displayed in the standard way, and to progressively compress the image nearer to the perimeter of the screen. Of course, one of the advantages of digitally simulating this 'lens-like' behaviour, is that we can instantly abandon it when it does not serve our purpose, or adopt a view which, for example, compresses only at two edges rather than all four.

Contents

Graphics Multimedia Virtual Environments Visualisation Contents