Abstract
Rationale. The accelerating nonlinearity revealed in contrast detection tasks must combine nonlinearities of all processing stages involved. If there were some means to disable these nonlinearities selectively, one could reveal the stages by linking them with the component nonlinearities. To exploit such an opportunity, the nonlinearities were probed by 1) presenting a contrast pedestal dichoptically and 2) introducing a luminance pedestal. We argue that in the first manipulation eliminates the nonlinearity at the binocular summation site and the second manipulation eliminates any nonlinearity in the luminance adaptation stage. Methods. The contrast target was a monocular Gaussian blob with sigma = 3 arc min. The matching pedestal in the other eye had a contrast of 10%; the luminance pedestal was a disk with 1 deg diameter presented upon a dark field. Results. We found that 1) monocular contrast detection upon a uniform field produces a quadratic (accelerating) nonlinearity; 2) dichoptic presentation of the contrast pedestal did not affect the quadratic nonlinearity; 3) presenting the test with the luminance pedestal eliminated the quadratic nonlinearity, revealing a hard threshold nonlinearity instead; 4) presentation of the dichoptic contrast pedestal with luminance pedestal in both eyes did not eliminate the threshold nonlinearity but halved its value. Conclusions. From results 1) and 2) we conclude that the quadratic nonlinearity is a property of the luminance adaptation mechanism in the retina. From results 3) and 4) we conclude that the threshold nonlinearity is located in the cortex and has two components: monocular and binocular, indicating that the monocular receptive fields of the binocular neurons are provided by the monocular cortical neurons, which, in turn, collect signals from the LGN. The stage where the critical noise is infused in the contrast signal must be located right after binocular summation.