Inside-out displays occasionally produce an interesting failure of spatial interpretation known as ``control reversals''. For reasons given in Chapter 3, it was thought that foreground occlusions might reduce this problem, resulting in a performance measure for the foreground occlusion effect. The literature on inside-out displays is summarized below.
The two simplest ways to represent the roll of an aircraft on a cockpit display are either: to keep a representation of the ground stationary and roll an aircraft icon; or to keep the aircraft icon stationary and roll the ground representation (see Figure 2.2). The former is referred to as an outside-in display; the latter, as an inside-out display. The terminology arises as follows. If one views a plane from above and behind (``outside''), a roll is seen as the plane moving with respect to the ground. Hence, a display in which the aircraft icon moves is called ``outside-in''. Conversely, if one sees the roll of the plane from inside looking out, what one sees with respect to the plane is the ground moving up either the left or right window of the cockpit. Hence, a display in which the ground representation moves is called ``inside-out''.
Despite the congruence between inside-out displays and the view out the window of the cockpit, inside-out displays are prone to a misinterpretation known as a ``control reversal''. When a control reversal occurs, the pilot inadvertently attempts to control the representation of the ground, rather than of the plane. For instance, in Figure 2.2 (A) we see a representation of a plane rolled with its right wing down. To correct for the roll, the pilot should bank the plane to the left. However, in a ``control reversal'' pilots will instead act as if they should correct for the roll by ``pushing down'' on the ground representation (i.e., banking the plane to the right).
Of course, pilots are well-trained to interpret inside-out displays correctly, and control reversals are rare. Nevertheless, Roscoe mentions that ``it is possible - even for airline pilots - to confuse the moving horizon bar of the gyroscopic attitude indicator and the fixed airplane symbol when they find themselves suddenly and unexpectedly in an unusual flight attitude.''  Roscoe suggests that this condition may be implicated in ``graveyard spirals'', which cause approximately 100 deaths per year in the U.S. in general aviation, as well as occasional commercial disasters, perhaps including the 1994 USAir 737 crash.
Given that the inside-out display is congruent to the view out the window of the cockpit, and that control reversals do not occur when looking out the window, why do control reversals occur when using inside-out displays?
There is a good discussion of this issue by Roscoe et al.  (p. 70):
The problem of pilot errors on moving-horizon attitude displays may be explained in the context of the psychological phenomenon of figure and ground. Although figure-ground definitions emphasize static aspects of the visual field, dynamic aspects dominate the flight situation. Psychologically, the part of the field of view that appears to be stationary is customarily called the background, and the object that is moving is called the figure. When the entire visual field moves in relation to the observer's eye, as occurs with head movement, the observer usually perceives correctly that the background is stationary.
The question then becomes: do the figure and ground relationships between the aircraft and the outside world change when the pilot shifts attention from the outside world to the attitude indicator on the panel inside the cockpit? If the pilot's frame of reference changes when a small, abstract instrument representation of the outside world is all that is available, as opposed to the outside world itself, this change must involve a shift in the figure-ground relationship. Specifically, the aircraft's instrument panel or even the framed aperture of an individual display face becomes the background against which the display elements move.
Roscoe et al.  (p. 70) mention an interesting technique for removing control reversals, which was the subject of Pilot Study AIIP2.
The highly resolved, dynamic, literal image in full color presented on a display screen by a projection-type flight periscope consistently yields the same stable figure-ground relationships for all pilots regardless of display size. With display screens subtending visual angles ranging from 30 degrees down to 7.5 degrees (a 2-inch screen viewed from 15 inches) and presenting a forward looking view as narrow as 3.75 degrees (x2 magnification), no control reversal was observed during more than 135 hours of flight experimentation involving more than 25 different pilots of widely varying experience.
Another method for reducing the control reversal problem is the ``Malcolm horizon'' , in which a light representing the horizon is extended across the cockpit. Aside from being easier to attend to than a small horizon display, this approach makes ``use of the neural programming which naturally orients us with the horizon.'' The authors discuss the importance of stimulating peripheral vision in order to achieve intuitive orientation with the horizon.