The contrast gain control of the cat retina.

This paper is about a phenomenon, and a mechanism to explain the phenomenon. The observation is that the dynamic transfer properties of the cat retina are altered when the contrast of the visual stimulus is changed. The peculiar dependence of the “contrast effect” on the spatial and temporal properties of the visual stimulus led us to the conclusion that the contrast effect is caused by a separate and distinct retinal mechanism. We call this mechanism the contrast gain control. Our experiments are performed on single fibers in the optic tract of cats anesthetized with urethane and paralyzed with gallamine and alloferin (cf. Hochstein and Shapley, 1976a for details). The visual stimulus is generated on a cathode ray tube (CRT) at a mean luminance of 20cd/m2. We classify the optic tract fibers as X or Y on the basis of their responses to contrast reversal of fine sinusoidal gratings. As spatial frequency of the grating is increased, the Y-cell response becomes predominantly nonlinear. The symptoms of this nonlinearity are that the cell’s response is mainly at twice the modulation frequency and is completely independent of the position, or spatial phase, of the grating stimulus (Enroth-Cugell and Robson, 1966; Hochstein and Shapley, 1976a). On the other hand, X-cells behave in a linear manner. They respond at the modulation frequency, and this fundamental response is a sinusoidal function of spatial phase. For classification, the modulation signal which controls the instantaneous value of the contrast of the grating pattern is a single sine-wave around 4 Hz. After the unit is classified, we change the temporal modulation signal to a sum of eight sinusoids of equal amplitude. These sinusoids are roughly a factor of 2 apart in frequency and span the range 0.2-31 Hz (cf. Victor, Shapley and Knight, 1977). When the experimenter looks at the CRT screen at a sine grating modulated by this sum of sinusoids, the grating appears to be varying in contrast in a random manner. If one listens to the response of an optic tract fiber to this visual stimulus, one hears its firing rate modulated in what seems a random manner. But the appearance of randomness is an illusion. One can recover the linear, or first order, response of the ganglion cell by Fourier analysis. One looks for Fourier components at the input temporal frequencies in the sum of sinusoids. One can also recover the best estimate of the nonlinear second order responses of the cell by looking at the Fourier components at sums and differences of the input frequencies (Victor et al., 1977). Second order responses will be discussed below. At first we wish to focus on the first order frequency response contained in the response amplitudes and phases at the indut frequencies. TWO first order frequency responses from a cat Y ganglion cell are shown in Fig. 1. In one experimental run the depth of modulation of the sinusoids was eight times higher than the other. The depth of the modulation is equivalent to the peak contrast the grating reaches. The contrast figures for the two frequency responses in Fig. 1 are 0.0125 peak contrast per sinusoid (root mean square (rms) average contrast of 0.025 for the sum of eight sinusoids) and 0.1 per sinusoid (rms average contrast of 0.2). Clearly, the frequency response is affected by contrast; at higher