Curvature coding in early visual system revealed by scale variance during adaptation to flashing circles

Abstract

How is curvature coded in the early human visual system? This question is the essence of basic shape recognition and may explain why humans are so successful in recognizing objects under difficult and varying environment conditions. The circle-polygon illusion produces polygonal percepts during adaptation when a static dark outline circle is pulsed at 2 Hz alternating with a gradient luminance circle. Both the radius and eccentricity of the stimulus were varied in a crossed design over 1-8 deg. Observers reported a circle or the polygon order and the strength of the percept. We test a lower level account that argues for curvature opponency between neurons against a higher level account that codes for whole shapes. This higher level account supports scale invariance, a property through which we recognize objects regardless of the object’s size on the retina. We show the following: (1) Scale invariance is not obeyed during adaptation. The mean order of the perceived polygon increased with stimulus size and decreased with eccentricity. It did not remain constant for a stimulus size. This also demonstrates that the curvature coding occurs in the early visual system. (2) We want to quantify the relationship parametrically between cortical size and polygon order. Linear regression analysis reveals that the cortical size of the stimulus is a better predictor of perceived polygon order. We computed the circumference of each stimulus in both retinal and cortical coordinates, taking cortical magnification into account and we extracted the slope that provides a metric for how well the polygon order reported fits the size and eccentricity of the stimulus, with its size adjusted to its retinal or cortical coordinates.