Color and intensity information representations by a network of turtle retinal ganglion cells

Visual stimuli are represented in neuronal firing of a 2D highly interconnected network of ganglion cells driven by the stimuli. As a result, the network is able to encode more features of the visual stimuli than can the single neurons that make up the network. We have recorded the simultaneous responses of 15 ganglion cells in the isolated, perfused turtle retina to simple visual stimuli circles of various diameters, intensities and colors. We trained a three-layer neural network to classify the responses in terms of stimulus color, size and intensity. The network performed poorly in the classification of the test stimuli. Further, concurrently recorded responses did not do significantly better at specifying the stimulus than did sequentially recorded responses. The approach provides a formalism that allows one to assess the effects of response variability on the ability of the network to classify responses