(A) Diagram of the weighted 4-quadrant model. Similar to ON/OFF processing in the visual system, the weighted 4-quadrant model splits the four differentially filtered signals into positive and negative components. As in the HRC, these component signals are paired, multiplied, and subtracted to produce four mirror anti-symmetric signals. We refer to these signals as HRC-quadrants. The model output is a weighted sum of the quadrant signals. We identify quadrants by whether they respond to the positive or negative components of each filtered signal and denote the four quadrants as (+ +), (+ −), (− +), and (− −). In this notation, the first index refers to the sign of the low-pass filtered signal (emanating from ), and the second refers to the high-pass filtered signal (emanating from ). (B) We measured the response of each quadrant to naturalistic motions and chose the quadrant weightings to minimize the mean squared error between the model output and the true velocity. (C) Comparison of the estimation performance of individual quadrants, multiple quadrants, and the HRC. The best two quadrants were (− −) and (− +); the best three also included (+ −). (D) The performance-optimized weighted 4-quadrant model reproduced the signs and approximate magnitudes of the psychophysical results.