Visual search experiments (Rensink & Cavanagh, 1993) have shown that rapid (preattentive) vision can interpret small regions of an image as cast shadows, provided that these regions are dark and lighting is assumed to be from above. Several experiments are presented that extend these results, mapping out the constraints used by this process in regards to color, texture, and the item casting the shadow.

Displays consisted of a set of vertically-oriented rectangles, with each rectangle having an attached region that could correspond to a cast shadow. Observers were asked to search for a target with a distinctive orientation to its attached region. In agreement with earlier work, when these regions were black and attached to the bottom of the rectangles (so that items corresponded to shadowed posts lit from above), search was slower than when the displays were rotated upside down (corresponding to lighting from below). This difference was not found when the attached regions were white, indicating that shadow interpretation did not occur for this condition.

Interpretation also did not occur when a dark region was outlined by a darker or lighter line, or had a dot along its boundary. It also did not occur when a texture was restricted to the region. However, results on blue and red regions and region outlines showed little evidence for purely chromatic constraints.

Interpretation occurred when the item casting the shadow was an outlined triangle and the attached region corresponded to a shadow cast by a rectangle, showing that geometrical constraints are not strong. But it did not occur when a gap was placed in the outline of a shadow caster, indicating that the process distinguishes between surface and line elements, with only the former considered capable of creating a shadow.