Motion sensitivity measured by a psychophysical linearizing technique.

Abstract

It has been proposed that the human visual system processes dynamically changing size nearly independently of other visual parameters, i.e., that there is a set of channels for changing size. A changing-size channel can be modeled as a sequence of linear and nonlinear stages. In particular, a pair of motion filters feed a changing-size filter that computes the algebraic difference between the outputs of the two motion filters. We report here that the peripheral motion-sensitive stage behaves approximately linearly with respect to both frequency and velocity. We measured the oscillation thresholds for 2-Hz antiphase (changing-size) oscillations of a test square before and after viewing different adapting oscillations. The adapting oscillation contained a constant 6-min-arc amplitude of antiphase 2-Hz oscillation in all sessions. However, in some sessions we added to the antiphase adapting oscillation an inphase component of frequency 2 Hz, whose effect was to render stationary two of the adapting square’s edges. We found that adding this 2-Hz inphase component reduced threshold elevation for the antiphase test stimulus, thus revealing an interaction between visual responses to the antiphase and inphase components of the adapting oscillation. Threshold elevations could be restored by adding a sufficient amount of a second “auxiliary” inphase oscillation to the adapting stimulus. In other words, a sufficiently large auxiliary oscillation amplitude eliminated the interaction between antiphase and inphase responses. This linearizing effect of the auxiliary oscillation depended on both the amplitude and the frequency f (in hertz) of the auxiliary oscillation. For all six values of f used, we found that altering frequency f was equivalent to a linear multiplying factor, as would be expected if the motion-sensitive first stage acted as a linear frequency filter. By playing off the amplitude versus the frequency f of the auxiliary oscillation so as to maintain a constant linearizing effect, we measured an attenuation characteristic for the first stage of the changing-size channel. This characteristic is flat from 1 to about 8 Hz and then rolls off at roughly 15 dB/octave.