Nonlinearities within the cat LGN cell receptive fields in simulated network with recurrent inhibition.

We investigated the receptive fields of principal cells from the cat's lateral geniculate nucleus cells. About 20% of the X type neurones showed clear nonlinearities of summation when stimulated by two simultaneously onset, small bars of light. The possible source of this nonlinearity was studied on a specially designed model of a one-layer neuronal network with inhibitory, recurrent interactions, intended to mimic the inhibitory influence exerted on geniculate relay cells by perigeniculate interneurones. The model, when activated from periphery by two stimuli-like input patterns, produced at the output side the nonsymmetrical profiles of the receptive fields sensitivity, similar to those obtained in real experiments. This nonlinear output appeared when some of the relay cells were inhibited below their firing level threshold and this effect was spread through the network by lateral inhibitory connections. It is concluded that physiologically observed nonlinearities of the order of single receptive field mechanisms can be predicted by a simple recurrent network.