Discrimination of Orientation-Defined Texture Edges

Michael S. Landy & S. Sabina Wolfson

Department of Psychology and Center for Neural Science, New York University, New York, NY

Invest Ophth & Vis Sci (1993), 34(4), abstract #584, page S819

Purpose. We examined factors governing texture segregation performance for edges defined solely by a change of orientation.

Methods. Each stimulus was a field of randomly placed blurred short line segments (length: .3 deg; 3.6 line segments/deg2; 100% white-on-gray Weber contrast; background luminance 30 cd/m2) viewed through a circular window (diam: 12.2 deg = 400 pixels; blurred edge). On the left side of the edge, line orientation was qL, and on the right side, qL + DqL. qL, DqL and edge orientation qE were varied. Subjects were required to discriminate between a straight edge and an edge with a sinusoidal profile (one cycle/12.2 deg; random phase; amplitude: .46 deg; 2I-2AFC). Absolute location of the edge in the circular window was randomized (+- 2.3 deg). Stimulus duration of each interval was 250 msec. Feedback was provided.

Results. For a vertical edge (qE = 0°), and for qL = 0° or 90°, best performance obtains when DqL = 90° (i.e. maximal orientation difference across the edge). However, for a vertical edge and diagonal line segments qL = 135°, better performance occurs at DqL = 45° and 135° than at DqL = 90°, although the latter is still better than chance. For qE = 45°, similar results were obtained: Rotating the entire stimulus 45° worsens performance across the board.

Conclusions. There are multiple determinants of texture segregation performance with orientation-defined edges: (1) the orientation difference across the edge (DqL), as would be predicted by most spatial channels models; (2) the orientations of the texture elements relative to the edge orientation (parallel and perpendicular are better than oblique), a configural effect that has been found before but is not predicted by channels-type models; and (3) an oblique effect for global texture edge orientation (qE) independent of local orientation.

None. Supported by NIH Grant EY08266, NASA RTOP 506-71-51 and NRC/NASA-ARC Research Associateship.