Contrast response of a movement-encoding system.

Abstract

The ability to identify the direction of apparent motion in a sequence of two short light pulses of different amplitudes at separate spatial locations was studied. The product of pulse amplitudes is a very poor predictor of such performance when one of the two signals is much higher in amplitude than the other: above a certain amplitude the probability of correct identification becomes virtually independent of the amplitude of the larger pulse. There was no noticeable difference in performance between low-high and high-low contrast sequences. Both the direction identification and the simple contrast-detection probabilities can be represented by the same psychometric function of the luminance increment delta L, provided that delta L is normalized by the nth power of the background luminance level, Lb. These results suggest that the general Reichardt-type scheme of movement encoding should be modified in the manner proposed for the fly's visual system [J. Opt. Soc. Am. A 6, 116 (1989)]: (1) the mean luminance is subtracted from the input signal before the signal is subjected to a nonlinear compression and (2) saturation characteristics are inserted into both branches of the two mirror-symmetric motion-detection subunits before multiplication of the input signals. The identical metric of the contrast response suggests that movement discrimination and luminance detection are two different special-purpose computations performed on the output of the same encoding network.