Abstract
Radial frequency (RF) contours can be used to represent globally processed shapes such as triangles (RF3), squares (RF4) and pentagons (RF5). It has previously been shown that more than one shape channel is needed to account for the detection of a broad-range of closed-contour shapes, however it remains unclear how many shape channels are required. Our first study used subthreshold summation to estimate the minimum number of shape channels necessary to explain performance for patterns up to RF10. To achieve this, we tested whether combining two RF components onto a single path (one of which is presented at half-threshold amplitude) improved detection thresholds for a single RF component, compared to thresholds for detecting that component in isolation. Threshold improvement occurred when the two shape components were the same RF but not when the components differed in RF, suggesting the operation of multiple channels, each with at most the width of a single RF. Channel sensitivity to luminance- and contrast-defined shape cues was then evaluated. A set of two experiments showed that, (1) positive and negative polarity luminance-defined contours mask each other and, luminance- and contrast-defined contours mask each other; (2) luminance and contrast cues can be assembled into a single global shape. The overall results show that the narrow-band channels which process global contour shapes effectively combine luminance- and contrast-defined form information.
Supported by ARC Grant # DP0666206 to DRB