Modeling texture-constancy of cortical grating cells

Cortical grating cells respond only and in a contrast-independent way to repetitive bar stimuli (bar gratings), but not to individual bars: their responses behave like the neural correlate of perceptual texture-constancy. We present a recurrent mean-field model for grating response. A first stage consists of idealized simple cell pools with identical preferred orientations, which are coupled by intracortical long-range patchy connections. In the second stage, grating cells sum up the simple cell input. The proposed model shows texture constancy, quantitatively reproduces the responses of macaque monkey grating cells and provides testable predictions of further cellular response properties.