The Resolution of Depth Ambiguity in the Kinetic Depth Effect

Abstract

The projected two-dimensional motions of a rotating object are sufficient to enable us to perceive its three-dimensional structure—a phenomenon called the kinetic depth effect (KDE). Although the object’s local, three-dimensional (3D) shape tends to be seen correctly, its global depth organisation, along with its perceived direction of rotation, often spontaneous reverses. My thesis is that these aspects of the KDE can be resolved, disambiguated, by adding temporal or spatial context.

In Experiments 2 to 4, I showed that the KDE could be disambiguated by adding temporal context. When I added the depth cues of contrast and superimposition they cooperated or competed in different ways in different individuals to disambiguate the KDE. I also discovered that the frequency of reversals and their time courses were best predicted from temporal context: the length of observation and an observer’s previous viewing history.

In Experiments 5 to 10, I showed that the KDE could be affected by, and could affect, unambiguous spatial contextual stimuli. When I added unambiguous surround stimuli to an ambiguous KDE object, I observed rotational linkage and rotational contrast. Linkage occurs when rotating objects share their axis of rotation. Contrast occurs when objects have different axes of rotation. These surround effects vary with speed, shape, rotation speed, and proximity. When I added an unambiguous surround stimulus that overlapped the ambiguous KDE object, I observed a phenomenon I call spreading concavity. In this, an unambiguous surround that by itself would be seen as a convex shape became concave when overlapping an ambiguous KDE cylinder.

The results of my experiments reveal the importance of temporal and spatial context for the KDE. Temporal context includes one’s history of viewing KDE stimuli and one’s history with different depth cues. These temporal contexts alter the frequency of reversals and can disambiguate the KDE. Spatial context has powerful effects on the KDE, disambiguating it, and, in the case of concavity spreading, supplanting alternative perceptual interpretation specified by depth cues present in the stimulus.

I discuss how context effects might be realized in terms of four hypotheses: Supplemented Physiology Hypothesis (SPH), Mechanical Constraints Hypothesis (MCH), Geometric Constraints Hypothesis (GCH), and Surface Reconstruction SFM Scheme (SRS).