Abstract
A fruitful method for investigating visual shape representation is to examine observers' sensitivity to different types of shape transformations. In previous work we found that, in two-part shapes, observers are more sensitive to changes in the orientation of the protruding part relative to the base shape, than they are to its translation along the base shape (Denisova et al., VSS 2009). One way to understand this result is that part articulation is a natural transformation in biological shapes, whereas part translation is not. The current study investigated the role of surface geometry by stereoscopically manipulating figure-ground assignment on the same set of contours. Observers viewed a near-vertical contour with a laterally protruding part, seen through a circular aperture. The disparity field was manipulated such that the same undulation was seen either as a protruding (positive) part or as an indentation (negative part). We measured observers' sensitivity to changes in orientation and location of the part. On each trial, observers viewed a standard shape, followed by two alternative shapes. One of the alternatives was identical to the standard; the other differed by a rotation or translation of the small part. Psychometric curves were thus derived for 2 transformations (rotation / translation) × 2 surface conditions (positive / negative part). For the orientation transformation, observers' sensitivity was on average higher for positive parts than for negative parts. However, for the location transformation, no systematic difference was observed between positive and negative parts. We note that, in the case of orientation change, the positive-part transformation occurs commonly in biological shapes, whereas the negative-part transformation does not. However, in the case of translation, both positive and negative-part transformations are unnatural. The visual system appears to be selectively tuned to shape transformations that occur naturally in biological shapes.
NSF DGE 0549115 (Rutgers IGERT in Perceptual Science) NSF CCF-0541185 NIMH T32 MH16434-31.