The threshold-quadratic nonlinearity appears to be a general property of signal integration in the recorded ganglion cells and presumably corresponds to the nonlinear processing that had been suggested to underlie Afatinib several

specific visual functions solved by the retina (Ölveczky et al., 2003, Gollisch and Meister, 2008, Gollisch and Meister, 2010 and Münch et al., 2009). Thresholding has been considered previously to lead to nonlinear receptive fields (Shapley and Victor, 1979, Victor and Shapley, 1979, Demb et al., 2001, Ölveczky et al., 2003, Geffen et al., 2007, Gollisch and Meister, 2008 and Münch et al., 2009), though often a threshold-linear operation has been hypothesized, rather than the threshold-quadratic transformation

observed in this study. Consistent with these previous findings, the source of this nonlinearity appears to be the bipolar cell input into the ganglion cell; the spatial scale of the nonlinearities GSK1210151A fits the receptive field size of bipolar cells (Figure 4), and this type of nonlinearity is not affected by a block of inhibitory neurotransmission (Figure 7). The threshold-quadratic nonlinearity may arise in the voltage response of individual bipolar cells (Burkhardt and Fahey, 1998) or in the synaptic transmission at the bipolar cell terminals (Baccus et al., 2008 and Molnar et al., 2009). It is noteworthy that iso-latency curves were more consistent in their shapes and always clearly displayed the quadratic part of the nonlinearity (Figure 3G), whereas iso-rate curves, even for cells that were not classified as homogeneity detectors, sometimes showed a tendency toward more linear integration (Figure 3H, see also Figure 3B for an example). This may be explained by local adaptation, for example, synaptic depression, which somewhat reduces the efficiency of strong local stimulation during the course of the spike burst. It is further interesting to note that no found linearly integrating ganglion cells were observed in our study.

This might be a feature of the investigated species; in the cat retina, for example, X-type cells would be predicted to have iso-response curves in the shape of straight lines. The particular sensitivity to homogeneous illumination of the receptive field in homogeneity detectors appears to arise from inhibitory interactions in the circuit. The nonconvex shape of the iso-rate curves was always abolished by removal of inhibition from the retinal circuitry, including experiments with reduced stimulus area so that different ranges of input into the system were tested. Otherwise, the nonconvex shape proved robust to changes in stimulus layout and overall activation level. Together with the success of the computational inhibition model, this supports a principal role of inhibition for generating the response features of homogeneity detectors.