Conclusion

Four fundamental points are made above:

• Both interreflections and shadows - either alone or in combination - can serve as cues for contact.

• The visual prominence of shadows and interreflections is not an indication of their effectiveness in conveying information about spatial organization.

• The presence of contact cues along the whole of the line of contact is more important than whether the cues involve shadows or indirect illumination.

• Crude approximations to shadows and interreflections are sufficient to establish a sense of contact, even when the subjective sense of realism is seriously compromised.

While the graphics community has known that shadows are useful, indirect illumination has often been considered only important for applications where high subjective realism is valued. This historical down-grading of indirect illumination relative to shadows is probably due to the difference in visual prominence, which we have argued is not directly related to their effectiveness as contact cues. Our results imply that even virtual reality systems where realism is not the primary goal might well benefit from at least approximating indirect lighting, particularly since in such cases simple techniques are likely to be sufficient for conveying an adequate sense of spatial organization. A further use is for visualization applications where demonstrating contact may be important, but shadows obscure too much information. The ability to use indirect lighting cues gives another choice to the designer of such applications.

Haddon and Forsyth, commenting on a computational analysis of the information available about scene geometry in shading patterns due to interreflections, observe that ``the best prospect for extracting shape information from shading is to construct programs that observe stylized properties of shading and associate those properties with shape primitives or their properties'' [20]. We reach the same conclusion approaching a closely related question from a human vision perspective: stylized patterns of lightness and darkness are sufficient to signal perceptions associated with shadows and interreflections in a way that is almost invariant to the actual radiance values that are present.

There are several important perceptual issues that are not addressed in this paper. We have demonstrated that in static scenes shadow and indirect illumination cues can establish object contact. When moving objects are brought into proximity with other objects, the dynamic cues may have different characteristics than the static cues. Also, non-diffuse effects may raise issues that do not arise in the diffuse case.

Finally, this paper is an example of the symbiotic relationship that is naturally arising between the graphics and psychology communities. Graphics researchers can provide variation to the optical behavior of the world which enable new sorts of programmable stimuli, and psychology researchers can help the graphics community prioritize what is rendered, so that efficient and effective algorithms can be developed.