From: A new analytical method for characterizing nonlinear visual processes with stimuli of arbitrary distribution: Theory and applications Journal of Vision. 2017;17(6):14. doi:10.1167/17.6.14 Figure Legend: (a) Plot of eigenvalues (absolute value) of the second-order kernel matrix. (b) The results of eigenvalue decomposition analysis of the second-order kernel matrix. The figures show the three largest and three smallest eigenvalues and the corresponding eigenvectors arranged in the frequency space. Figures in the top row are eigenvector plots for positive eigenvalues, showing the combinations of gratings that facilitate the upward response, whereas the figures in the bottom row are those for negative eigenvalues that facilitate the downward response. The top left plot for the largest positive eigenvalue indicates that low spatial frequency gratings moving upward and high spatial frequency gratings moving downward work together to facilitate an assessment of upward motion, whereas high spatial frequency gratings moving upward are negatively related with such effects. The bottom left plot for the smallest negative eigenvalue is a horizontally flipped version of the pattern in the top left plot. Date of download: 12/19/2017 The Association for Research in Vision and Ophthalmology Copyright © 2017. All rights reserved.