Saccadic localization in the presence of cues to three-dimensional shape

Saccades directed to simple two-dimensional (2D) target shapes under instructions to look at the target as a whole land near the center of gravity (COG) of the shape with a high degree of precision (He & Kowler, 1991; Kowler & Blaser, 1995; McGowan, Kowler, Sharma, & Chubb, 1998; Melcher & Kowler, 1999; Vishwanath, Kowler, & Feldman, 2000). This pattern of performance has been attributed to the averaging of visual signals across the shape. Natural objects, however, are three-dimensional (3D), and the shape of the object can differ dramatically from its 2D retinal projection. This study examined saccadic localization of computer-generated perspective images of 3D shapes. Targets were made to appear either 2D or 3D by manipulating shading, context, and contour cues. Average saccadic landing positions (SD approximately 10% eccentricity) fell at either the 2D or 3D COG, and occasionally in between, depending on the nature of the 3D cues and the subject. The results show that saccades directed to objects are not compelled to land at the 2D COG, but can be sensitive to other visual cues, such as cues to 3D structure. One way to account for these results, without abandoning the averaging mechanism that has accounted well for performance with simple 2D shapes, is for saccadic landing position to be computed based on averaging across a weighted representation of the shape in which portions projected to be located at a greater distance receive more weight.