The Importance of Foreground Occlusions

The importance of the visual background to determining the selected rest frame was discussed above. In accordance with the presence hypothesis, one would expect that the visual background would also play a role in determining the sense of presence.

It is known that increasing the FOV of a display, everything else being equal, increases the sense of presence. This is consistent with the above argument: as the FOV of a display is increased, it absorbs more of the visual field, which should increase the likelihood that the contents of the display will be interpreted as defining the visual background. If the contents of the display are interpreted as defining the visual background, they will influence the selected rest frame; this should in turn increase the sense of presence.

The importance of FOV is well-known to the virtual environments community. Less well-known is the importance of a foreground occlusion (see Section 2.4.1). Because a foreground occlusion places the display at a greater distance than all other visual cues, it becomes more likely that the contents of the display will be interpreted as defining the visual background.

One might consequently predict that a foreground occlusion will influence the selected rest frame and (therefore) increase reported presence. Two sequences of experiments on this question are reported in this dissertation. The first sequence examined the effect of foreground occlusions on reported presence.

The second sequence, adapted from the problem of control reversals in inside-out displays (see Section 2.4.3), sought a performance measure related to the influence of a foreground occlusion on the selected rest frame. Roscoe et al.'s ``figure and ground'' discussion of the control reversal problem has a natural expression in terms of the RFC. In order for an inside-out display to be interpreted correctly, a difference in the orientation of the ground representation and the aircraft icon has to be interpreted as a roll of the aircraft icon, not of the ground representation. That is, the ground representation has to define the selected rest frame. This happens automatically when looking out the window of the real aircraft: the real ground defines the visual background, and hence the selected rest frame, and hence a change in the roll angle between the real ground and the real aircraft is interpreted as a roll of the real aircraft.

The interpretation of a roll depicted on an inside-out display is much less intuitive than a roll seen out of a real cockpit window. The ground representation on the inside-out display does not have a wide FOV and does not define the visual background. The selected rest frame tends to be determined by the cockpit as a whole. Since the aircraft icon on the inside-out display is fixed with respect to the cockpit, a change in the roll angle between the ground representation of the inside-out display and the aircraft icon is interpreted as a roll of the ground representation. In the absence of training or under stressful conditions, this can lead to control reversals.

In view of this line of thinking, it is interesting to consider the two remedies for control reversals discussed in Section 2.4.3. The use of a ``highly resolved, dynamic, literal image in full color presented on a display screen'' is consistent with factors known to increase the sense of presence (see Section 2.2). The presence hypothesis suggests the following: the ``literal image'' increases the sense of presence, which equivalently maps the selected rest frame into the natural reference frame for the display (the ground representation), which consequently reduces control reversals.

The second remedy discussed was the Malcolm horizon, which extends the ground representation of the inside-out display across the cockpit. In part, the value of the Malcolm horizon arises from providing a wide FOV stimulus which makes it easier to detect small angular motions. However, recent research on narrow FOV vection (see Section 2.4.1) shows that motion perception depends heavily on what one perceives to be the visual background. It therefore seems plausible that the Malcolm horizon also has a more indirect influence: expanding the FOV increases the chance that the ground representation will be interpreted as the visual background, and hence that the ground representation will influence the selected rest frame.

More generally, one might expect that any manipulation which increases the sense that the ground representation of the inside-out display is in the visual background would serve to reduce control reversals. One such manipulation is to provide a foreground occlusion. In view of the Malcolm horizon, and in view of the above discussion of the parallel between increasing the FOV and providing a foreground occlusion, one might predict that placing a foreground occlusion in front of an inside-out display would serve to reduce control reversals.

This idea is hardly of practical value for real cockpits, since a foreground occlusion which blocked out other visual cues at the same or greater distance as the inside-out display (i.e., the rest of the cockpit) would not be useful. Whether a foreground occlusion can in principle reduce control reversals is an interesting question, however, for at least two reasons.

1.

It provides a possible performance measure for the foreground occlusion effect.

2.

It is a test of the theory that the perceived visual background is a determining factor in the occurrence of control reversals. This might be of relevance to the design of virtual cockpits presented on a display, in which the ground representation could be put at a greater distance than the rest of the virtual cockpit using stereoscopic, accommodative, or other depth cues.

Just as the frequency of control reversals is a potential measure for the foreground occlusion effect, it is also a potential measure for any other presence manipulation. This suggests that the effect of a wide variety of display manipulations on presence could be measured by evaluating the frequency of control reversals.

Human Interface Technology Lab

Human Interface Technology Lab