The Determination of Inertial Cross-Over Amplitudes

The procedure described here determines the relative influence of visual and inertial stimuli on the sense of self-motion. See Figure 3.1, on page , for a diagram illustrating the procedure.

Both inertial (chair) and visual (HMD) oscillations were at .1 Hz, but the phases conflicted. The degree to which participants identified with the visual vs. inertial self-motion cues was determined by having them indicate with a toggle switch the perceived right and left extremes of the inertial motion. If the inertial amplitude is sufficiently low, one tends to inadvertently indicate the vection cues, rather than the inertial motion of the chair (even though, with one's eyes closed, one could correctly follow the chair motion at the same inertial amplitude). The term ``inertial dominance'' will be used when the observer correctly indicates the inertial motion, and the term ``visual dominance'' when the observer indicates the vection stimulus, despite attempting to signal the chair motion.

Using the PEST procedure (see Section 3.8.4) to adjust the inertial amplitude after each trial, the inertial amplitude at which the participant switched between visual and inertial dominance was determined. This is termed the ``cross-over amplitude''. The presence hypothesis implies that the cross-over amplitude should be a presence measure. The research described below investigated this measure.

The experiments described in this chapter ran trials of two visual conditions in parallel in each session. To counter-balance order effects, trials from the two conditions followed an ABBA pattern within each session. In addition, the condition with the first trial was counter-balanced across participants, and within participants across sessions.

The visual amplitude for all trials was 30$^{\circ}$/sec. This was picked for saliency after informal trials. The experiment began with an inertial amplitude of 15$^{\circ}$/sec. The initial step-size for the PEST procedure was 10$^{\circ}$/sec. The termination condition was a step-size of at most 5$^{\circ}$/sec. This value were picked to converge fairly quickly, in order to keep sessions in the 1 hour range, counting introductions, administering the Embedded Figures Test, simulator sickness questionnaire data collection, etc.

The cross-over data reported below are the midpoints of the range to which the inertial amplitude was narrowed in $^{\circ}$/sec. This is half the maximum peak-to-peak velocity difference over one cycle. See above for a discussion of these units.

The phase difference between the inertial and vection cues was always $\pm90^{\circ}$. A $90^{\circ}$ magnitude phase difference is useful in that the two curves are sufficiently far apart that the difference between visual and inertial dominance is readily apparent in the data, but not so far apart that the vection and inertial self-motion cues become clearly distinct. A magnitude of $90^{\circ}$ was picked on the basis of informal trials.

In Pilot Study AIP3, a random choice of either $\pm90^{\circ}$ phase angle was used, to avoid a possible learning effect which might occur from holding the phase angle constant. This resulted in a poor test-retest correlation (.38), possibly due to a systematic difference in the difficulty of the two phase angles.

Pilot Study AIP4 used the same vection-inertial phase angle of $+90^{\circ}$for all trials. For a small participant sample (n = 4), this resulted in a test-retest correlation of .99. There was not a within-subjects drop in the cross-over amplitude from the first to the second session. All 4 of the participants believed (incorrectly) that the phase angle was changing across trials. This suggested that a learning effect might not be a major issue with a single phase angle.

Each participant was given an initial practice trial with the HMD turned off at an inertial amplitude of 10$^{\circ}$/sec. This served both to introduce the participant to the procedures and to check their inertial motion detection. Only one participant was screened out on the basis of this test.

Between trials, while the chair was at rest, participants were asked to relax and close their eyes. Before each new trial, after starting the chair in motion, participants were asked to count down by 7's with their eyes closed for 25 seconds from an arbitrary number selected by the experimenter. This provided a distraction which prevented the participant from ``locking in'' to the inertial motion. They then continued to count down for an additional 25 seconds with their eyes open. This gave the vection cues time to build up an effect. Next, participants were asked to stop counting and start signaling the perceived endpoints of the chair's inertial motion while attending to the visual scene.

In written and oral instructions, participants were asked to

1.

Pay attention to the visual scene when your eyes are open.

2.

Signal how you think the chair is physically moving.

3.

Close your eyes after each trial until asked to open them.

A trial always ran for at least 4 signals. Often the trial was halted after four signals, as all four would indicate either clear visual or clear inertial dominance. If not, approximately 30 seconds to a minute of signals would be recorded to examine whether a clear trend developed. The trial would then be halted and the average phase distance of the signals from the inertial and vection stimuli would be recorded. After the trial, the chair was brought smoothly down to zero velocity.

At inertial amplitudes markedly below or above the cross-over amplitude, participants signaled perceived endpoints of their inertial motion which were roughly a phase distance of 10$^{\circ}$ from either the vection or inertial cues, depending on whether the participant was in inertial or visual dominance. As the cross-over amplitude was approached, the phase distances equalized in a predictable if somewhat haphazard manner. This raised the issue of what decision to make for the PEST procedure if the phase angles between the participant's signals and the two motion stimuli were indistinguishable (defined to mean: within 10$^{\circ}$ of each other). The decision made was as follows: if the cross-over amplitude had already been determined to within a 5$^{\circ}$/sec range, simply take the current amplitude as the cross-over amplitude. If not, retest.

Human Interface Technology Lab