Linear extrapolation does not explain the mislocalization of a triangle's occluded vertex

Poster Presentation: Monday, May 19, 2025, 8:30 am – 12:30 pm, Pavilion
Session: Perceptual Organization: Parts, wholes, shapes and objects

Tess White1, Jacob Tack1, Gideon Caplovitz1; 1University of Nevada, Reno

In previous year’s VSS conferences, we presented a variety of stimulus configurations in which the perceived intersection point of two converging lines is systematically biased away from the true intersection point. We have found a consistent pattern of result for configurations that include partially occluded triangles, partial triangles without explicit occluders and even pairs of converging lines. The bias persists under both binocular and monocular viewing and is modulated by both the length of the visible lines and the convergence angle, with greater biases observed for more acute angles. In triangle configurations, observers underestimate the inferred vertex (closer to the base than it really is). While this contraction is somewhat consistent with interpolation models of contour continuation in amodal completion, such models are based on smooth-curvilinear interpolation and do not readily account for completing a singular vertex or intersection point. Rather than interpolation, our current research seeks to determine whether the contraction is a result of an extrapolation mechanism. Across two experiments our observers were presented with linear line segments matched in length and orientation to conditions used in our previous experiments and performed a linear extrapolation task in which they place a dot on the other end of the screen where they thought the line segment would extrapolate to. In both Experiments, we found systematic biases in how observers extrapolate with a pattern of results that conflict with our previous findings. For example, acute line angles relative to vertical tend to be pulled towards the vertical, a finding that is consistent with expansion not contraction. Unlike the results of our previous studies, two-line extrapolation results in the strongest effects in obtuse angle conditions (120°). Based on these data, we conclude that linear extrapolation does not explain the systematic bias of a triangle’s vertex or two converging lines’ intersection point.