Abstract
The stereo matching direction is underdetermined for gratings and other 1-D patterns—the ‘aperture effect’—but can range over 360° for 2-D patterns, such as plaids. How does perceived depth vary with disparity direction? To find an answer, we measured the perceived depth between a grating and a plaid and between two plaids. The results show that two relative disparity signals underlie perceived depth. A central grating or plaid bounded by a circular Gaussian window, and a surrounding plaid bounded by an annular Gaussian window, appeared simultaneously for 150 ms. We fixed the disparity direction and magnitude of the surrounding plaid and varied the disparity magnitude of the central grating or plaid across trials. Subjects judged the central pattern as ‘near’ or ‘far’ relative to the surrounding plaid. We measured the disparity of the center carrier yielding a perceived depth match between center and surround as a function of the orientation of the central grating and the disparity direction of the central plaid. A grating matched the plaid in depth when the grating had a disparity equal to the projection of the plaid's disparity vector onto the disparity axis perpendicular to the grating's orientation; this is the disparity the grating would have had it been a component of the plaid. The depth match was unaffected by the relative horizontal disparity of the two stimuli. By contrast, two plaids were perceived to match in depth when their disparities had equal horizontal components, regardless of any vertical disparity difference within at least ±60° of horizontal. Perceived depth has been thought to depend on horizontal disparity or on the horizontal:vertical disparity ratio. Our results show that neither of these rules is general. However, a modified intersection-of-constraints computation accounts for all the data.