Abstract
Problem: Some geometric constraints on contour interpolation have been shown to be largely scale-invariant (Banton & Levi, 1992; Shipley & Kellman, 1992). Kellman & Shipley (1991) hypothesized that the constraints of contour relatability may include a small tolerance for misaligned parallel edges that is not scale-invariant. Earlier we reported data suggesting that this tolerance is constant in retinal terms, rather than scale-invariant (Hilger & Kellman, 2005). Here we report more comprehensive studies of the relation of contour interpolation to retinal misalignment. We used different stimulus types (step-edges vs. Gabors) and misalignment methods (element misalignment vs. phase misalignment). Methods: Tolerance for misalignment was tested in a two-interval forced-choice path detection paradigm (Field, Hayes & Hess, 1993). Targets were paths of nine spatially separated contour segments that were collinear or misaligned to varying degrees relative to the axis of global path alignment. Paths were presented in noise consisting of 247 identical contour segments, randomly oriented. The amount of retinal misalignment was manipulated within subjects while the inducer type (step-edge or Gabor) and misalignment type (element misalignment or phase misalignment) were varied between subjects. Results: Element-misaligned stimuli largely confirmed a retinal tolerance, with interpolation effects disappearing by about 16 min. Both step-edge and Gabor stimuli showed similar absolute performance levels and similar decreases with increasing misalignment, falling to chance between 12–16 min. In contrast, phase-misaligned Gabors (with aligned envelopes) showed superior performance overall and a slower decline in performance with increasing misalignment compared to element-misaligned stimuli. Conclusions: Tolerance for misalignment in the interpolation of edges is determined by a retinal metric, decreasing gradually up to 12–16 min. This tolerance is largely identical for both illusory edges and Gabor-based paths. Other aspects of an array, such as perfectly aligned envelopes, may increase this tolerance in a search-based task.
Supported by National Eye Institute EY13518 to PJK.