Introduction

This chapter addresses the ``presence hypothesis'' from Section 3.3.3 and its application to the measurement of presence. The ``presence hypothesis'' claims that the degree to which the rest frame implied by a virtual environment perceptually overwhelms a conflicting rest frame implied by the real (external) environment should constitute a presence measure.

The following questions are addressed in this chapter.

1.

How should a rest frame conflict measure be implemented?

2.

Is the presence hypothesis correct? That is, does a measure based on real-virtual rest frame conflict evaluate our subjective sense of presence?

3.

What is the test-retest correlation of a rest frame conflict measure?

4.

How is the rest frame conflict measure related to field dependency? Presence and field dependency both measure the degree to which one is drawn into a visual scene. It might therefore be expected of a good presence measure that between-subject variation would correlate with between-subject variation in field dependency.

In principle, many real-virtual rest frame conflict measures are possible. See Section 3.3.2 for a classification of spatial ``illusions'' which can be induced by visual stimuli. Any of these can be harnessed to create a rest frame conflict measure, by implying a rest frame with the virtual cues which conflicts with the rest frame implied by the real cues. The rest frame conflict measure then records the degree to which the observer's spatial perception is dominated by the virtual, rather than real, cues.

A straight-forward rest frame conflict measure can be based on altering the perception of the vertical. Visual cues from a display may be set to indicate the vertical to be different from gravity. Measuring the perception of the vertical, for instance by having the observer adjust a rod to indicate the perceived vertical, measures the degree to which the visual rest frame has overwhelmed the inertial rest frame.

While rest frame conflict measures based on perceived orientation are quite possible, they face the difficulty that the inertial perception of gravity is very strong. This tends to make the influence of the visual scene difficult to measure, particularly for virtual environments which are not very compelling. The research reported in this chapter therefore focused on a measure in which gravity is not a factor: conflicting inertial and visual self-motion cues in the horizontal plane. Given conflicting inertial and visual self-motion cues, one's perception of self-motion may be determined by the inertial cues, by the visual cues, or by some combination. As the inertial amplitude is lowered, it becomes increasingly likely that the visual cues will determine the perception of self-motion.

The approach was to develop a ``visual-inertial nulling measure'' for finding the inertial amplitude at which participants cross-over between inertial and visual dominance of the perception of self-motion. The hypothesis was that a visual condition which was more ``presence inducing'' would require stronger inertial cues to match the visual cues perceptually, and that this would be reflected in a higher visual-inertial cross-over amplitude.

The technique developed here is intended primarily as a proof-of-principle: to establish a rest frame conflict measure which factors out gravity, and to examine its relationship to a reported presence measure and the test-retest correlation of both measures. The measure used in this chapter is not a final answer to the need for sensitive and convenient presence measures. In particular, it is cumbersome (it requires the full attention of the participant to perform the measurement task) and is not suited to interactive virtual environments (since the time-course of both the visual and inertial motion cues have to be tightly controlled by the experimenter). Pilot Study AIIIP2 (see Appendix D) raised the possibility of a more graceful rest frame conflict measure, based on the induced motion of a background grid. See Chapter 8 for a discussion of this possibility.

Two experiments are described in this chapter. The first experiment, AIE1, examined three FOV's: 16$^{\circ}$, 32$^{\circ}$ and 48$^{\circ}$FOV. It was predicted that wider FOV would result in higher presence on both a self-report and visual-inertial nulling measure. The second experiment, AIE2, focused on a meaningful/random visual scene manipulation. The random condition was created by randomizing the pixels in the meaningful scene. The meaningful/random manipulation appeared more likely to result in a clear main effect on both measures than the FOV manipulation described above. The main effect was predicted to indicate higher presence on both measures in the meaningful condition.

The meaningful/random manipulation reported here is similar to the meaningful/meaningless manipulation described in Appendix A. The meaningful/meaningless manipulation compared meaningful with matched meaningless chess positions. This found a difference in reported presence on a purely cognitive variable, in favor of higher presence for the meaningful condition.

The meaningful/random manipulation described in Experiment AIE2 is not a purely cognitive manipulation. In addition to affecting level of meaning, it also affects low-level perceptual factors such as number and distribution of edges. Thus, while the results of Experiment AIE2 could be taken as supporting the conclusion of Appendix A that meaning affects presence, it does not do so unambiguously.

Human Interface Technology Lab