Linear and non-linear properties of simple and complex receptive fields in area 17 of the cat visual cortex. A model of the field.

Abstract

Responses of neurones in area 17 of the cat visual cortex to gratings with sinusoidal distribution of luminance were studied. Impulse responses of the simple fields were found to be modulated by gratings passing the field. The complex fields proved to be of three types: with modulated responses, unmodulated responses, and with modulated responses against unmodulated background. Amplitudephase characteristic measured were inverse Fourier transformed to obtain the field's weighting function. Simultaneously the amplitude-phase characteristic was reconstructed from the weighting function from the responses to edges and bars, with the use of the Fourier transform. Cross-comparison of the reconstructed amplitudephase characteristics and the weighting functions showed that a receptive field has some linear properties but, strictly considered, is a non-linear system. Simple fields display the largest degree of linearity. The weighting function of a simple field does practically not depend on the stimulus velocity. At lowest velocities, however, there additional waves in the weighting function can appear, which results in a narrower bandwidth. The higher the complexity of the field that is to say the greater the number of subfields of the preceding level are, the greater are the deviations from linearity until it is completely lost in some complex fields with the unmodulated type of response. As linear methods are inadequate for dealing with cortical receptive fields, their qualitative identification was performed in model experiments on a computer. A model was fitted to observations by varying the parameters of the model for a receptive field, such as the number and weights of subfields, the extent of their spatial overlap, the time constant of excitatory and inhibitory processes and detection level. The evidence obtained suggest that the receptive fields of the visual cortex form a system of operators which perform the expansion of the image on non-classical pattern. Such an expansion can be termed a piecewise quasi-Fourierdescription. Although receptive fields of the visual cortex have been studied in a great number of investigations, the properties of the fields as described by linear analysis are as yet imperfectly understood. In the present paper we have studied responses of simple and complex receptive fields in area 17 of the cat visual cortex to moving gratings of different spacial frequency with sinusoidal distribution of luminance. The amplitudephase characteristics obtained were inverse Fourier transformed, and the weighting functions reconstructed were compared with the weighting functions obtrined from responses to bars or edges moved through the field. In a similar manner, the weighting function obtained from the responses to bars or edges were Fourier transformed and the corresponding frequency characteristics were compared with those experimentally obtained. Such cross-comparison allowed us to make estimates of linear and non-linear properties of a field to reveal sources of non-linearity. Based on these results we propose a model of the field and discuss a possible role of the field in processing of visual information.