Abstract

We have previously shown¹ that observers can accurately detect changes in the relative position of a vertical intersector on a horizontal line segment. Performance is not seriously compromised by varying test pattern size relative to that of the reference or by rotating the entire test figure. Here we report the effects of either reflecting the test pattern left-right around its midpoint, inverting it, or rotating it by 90°. Despite marked changes in pattern appearance, neither manipulation significantly degrades observers' ability to make the required judgment. A sufficient solution to each of these tasks would be to estimate and compare the lengths of the appropriate line segments. To determine whether this is the necessary strategy, we utilized a test figure in which the intersection is represented by a significant gap centered around the intercept. Since equal absolute amounts are removed from each side, the ratios of line segment lengths are no longer comparable. In addition, the geometric cues provided by a 2-D intersector are no longer available. Performance on this task is surprisingly good. These data imply that simple perceptual models may not be adequate to account for the results.

© 1988 Optical Society of America