A Configural Shape Illusion
Stephen E. Palmer, Karen Schloss, & Francesca Fortenbaugh

A new illusion -- the configural shape illusion (CSI) -- is reported in which the shape of a rectangle is systematically distorted by an attached/adjacent contextual region, when both are seen as part of a single configuration. In particular, the rectangle's perceived aspect ratio changes in a direction consistent with the aspect ratio of the whole configuration. We measured the magnitude of this illusion in two ways. First, observers adjusted the height and width of a separate, unattached rectangle to match those dimensions of a rectangle that was part of various configurations. Second, observers adjusted the height and width of the rectangle within various configurations to appear perfectly square. Systematic CSIs were present using both procedures, but their magnitude depended on the spatial and color relations between the rectangle and the adjacent context. The results are consistent with the hypothesis that the illusion is greater to the extent that the elements in the configuration are strongly related by virtue of standard grouping factors, including connectedness, proximity, good continuation, lightness similarity, hue similarity, and shape similarity. Somewhat surprisingly, the illusion was stronger when the contextual region was smaller, suggesting that the magnitude of the illusion may be governed more by the proportion of the entire configuration occupied by the target rectangle than by the aspect ratio of the entire configuration itself. Similar effects are apparent for the aspect ratio of an oval, although the distortions are less pronounced. The relation between the CSI and the occlusion illusion (Kanizsa, 1979; Palmer, Brooks & Lai, 2007) will also be discussed.

Extremal Edges: A powerful cue to Depth & Figure-Ground Organization
Stephen Palmer and Tandra Ghose

Edges arising from depth discontinuities are powerful cues to figure-ground organization (FGO) in 2-D images. We studied pychophysically whether surface convexity and extremal edges (EEs) are effective cues to depth and FGO. EEs are viewpoint-specific tangent points of self-occlusion on smoothly curved, convex surfaces. A simple ecological analysis of viewpoint constraints shows that the curved surface producing an EE is likely to be closer to the observer than the surface on the other, non-EE side of the contour.Four experiments examined whether EEs and 3-D surface convexity operate as strong cues to depth and figure-ground perception. Experiment 1 used simple luminance profiles (e.g., the positive half of a sinusoid) to simulate shading gradients in simple bipartite displays. The results showed that observers are very likely to perceive both convex surfaces and EEs as closer and figural, but EEs are more potent than surface convexity alone. Experiment 2 showed similar effects when EEs were rendered via texture gradients of checkerboard surfaces that contain neither shading and nor implicit T-junctions. Experiment 3 used shading gradients in ray-traced images of surfaces of revolution to study the effects of EEs versus other individual figure-ground cues (region size, 2-D edge convexity, surroundedness, and familiarity). The results show that EEs dominate all of these factors. Experiment 4 used ray-traced images of EE shading patterns on simpler convex surfaces ("pillows") to control for certain shading artifacts in Experiment 3. EE effects still dominated the other cues studied (2-D edge convexity, size, and their combination). The data clearly demonstrate that extremal edges are among the most powerful cues to depth across a contour and to figure-ground organization.

Extremal Edge Project details

VSS 2005 Poster

The Occlusion Illusion
Stephen Palmer, Kevin Lai, and Joseph Brooks

We investigated a size illusion first reported by Kanizsa (1979) in which a figure bounded by an occluding edge looks larger than the same figure not bounded by an occluder. We call this phenomenon an occlusion illusion. We conducted a series of experiments to document the magnitude of this illusion and distinguish between two hypotheses about its perceptual basis. The size-distance relation (Emmert's Law) predicts that the farther of two objects with equally-sized retinal projections should be seen as larger. Because occlusion implies relative distance order, the size-distance relation serves as one explanation of the occlusion illusion. Alternatively, the illusion may reflect an extension of the occluded surface at the occluding border without a uniform increase in the perceived size of the figure. The distinctive prediction between these two theories is whether the occlusion illusion reflects a change in the overall size of the figure (size-distance theory) or a shape change in which the occluded edges perceptually extend beyond their intersection points with the occluder. This extension would form a greater area of a slightly different shape. To test this, we first used a staircase procedure to measure the magnitude of the occlusion illusion under various occlusion conditions and various occluding and occluded shapes. We found that the strength of the illusion varied with the strength of the cue to occlusion. In a second experiment we specifically addressed the two hypotheses about the origin of the illusion. In the first phase of this experiment each participant completed two staircase procedures. One staircase measured the point of subjective equality (PSE) for the overall size of the figure in comparison to the partially-occluded figure. The other staircase measured the PSE for extension of the edges beyond their intersection points with the occluder. The two figures representing the PSEs for each of these procedures were then compared to the partially occluded figure showing the illusion. Participants were significantly more likely to judge the figure with the extended edges as more similar to the example of the illusion. This suggests that the occlusion illusion includes a modal extension of edges that intersect with an occluding object.

See a version of the occlusion illusion

VSS 2004 Poster

Grouping Occurs Both Before and After Constancy
Stephen Palmer and Joseph Brooks

Previous results from our laboratory have shown that perceptual grouping occurs after various kinds of constancy processing, because perceived grouping can be strongly influenced by depth, post-constancy lightness perception, amodal completion, and illusory contours. We now report results showing that grouping also occurs before constancy processing. The general logic is to demonstrate that grouping both affects constancy and is affected by constancy. In shape constancy, for example, we first demonstrate that pictorial depth cues influence whether observers see a central column of ambiguous ovals as grouping with circles in the frontal plane or with ovals in the frontal plane. We then also show that whether observers perceive an ambiguous oval as a circle slanted in depth or as an oval in the frontal plane is strongly influenced by whether it is grouped with a surrounding trapezoid (consistent with the circular interpretation) or with a surrounding square (consistent with the oval interpretation). Proximity, common fate, and color similarity all have strong effects in our displays. We have also demonstrated similar kinds of grouping effects on edge assignment in depth perception for textured surfaces: the edge is assigned to the region whose texture elements group with the edge according to factors such as common fate, proximity, color similarity, and orientational similarity. Analogous grouping effects also appear to occur in displays involving lightness constancy. These results are inconsistent with aspects of Palmer and Rock's (1994) theory of perceptual organization, but consistent with an alternative formulation in which grouping occurs throughout perceptual processing.

See the Grouping and Constancy Demos

Figure-Ground Organization and Edge-Texture Grouping
Stephen Palmer and Joseph Brooks

A new class of information about depth across an edge and figure-ground organization - edge texture grouping - is described and demonstrated. The central issue in both cases is whether the edge between two regions "belongs to" (or "is grouped with") one side or the other. The side that is grouped with the edge should be perceived as closer and shaped by the edge, whereas the ungrouped side should be perceived as farther away and extending (unshaped) behind the edge. This analysis implies that classic grouping factors that relate various edge properties to corresponding textural properties within adjacent regions should systematically influence perceived depth across an edge and figural status. We report the results of several experiments that strongly support this claim for the grouping factors of common fate, proximity, synchrony, and similarity of blur, color, and orientation. Further, we argue that evidence of grouping in certain moving and flickering displays that are ecologically unnatural suggest that these effects are mediated by grouping processes rather than by inferences based on simple ecological statistics. The edge-texture grouping hypothesis provides a unified account of the present phenomena and others previously reported in the depth perception literature (e.g., Yonas, Craton & Thompson, 1987) as well as a coherent explanation of Weisstein's anomalous findings about the effects of spatial frequency and flicker on depth and figural status (e.g., Klymenko & Weisstein, 1986). Edge-texture grouping effects have important implications for depth and shape algorithms in computer vision as well as for corresponding perceptual and neural processes in human vision.

See the Figure-Ground and Grouping Demos