A possible source is narrow-field amacrine cells (Masland, 2001 and Chen et al., 2010). The iso-latency MDV3100 cost curves were not affected by the inhibitory mechanism that underlies the gain control; iso-latency curves always depicted the standard threshold-quadratic nonlinearity. This can simply be explained by a temporal delay of inhibition, resulting from involvement of an additional synaptic stage as compared to the direct excitation from bipolar cells (Werblin and Dowling, 1969, Roska et al., 2006 and Cafaro and Rieke, 2010). Note that the inhibition may act as a

direct input into the ganglion cells or indirectly by suppressing or modulating the bipolar cell output; the functional characterizations of the present study do not distinguish between these circuit features. The inhibition makes strong local stimuli that involve only a subset of available bipolar cells OSI-744 cost less effective, or in other words, it creates a particular sensitivity for spatially homogeneous stimulation when the activity load is shared between all available bipolar cells with weaker individual activation. The characteristic notch in the iso-rate curves of homogeneity detectors can thus be explained completely by local sensitivity changes without the need to evoke a direct interaction between the subregions of the receptive field. Interestingly, an example for the required strongly nonlinear activation of inhibition

has recently been found in paired recordings of certain amacrine cells and their presynaptic bipolar cells (Jarsky et al., 2011). The disproportionally stronger activation of inhibition for stronger stimulation also explains why the iso-rate curve

shapes differ for different target spike counts. The effect of inhibition becomes stronger with stronger stimulation, and consequently the notch in the iso-rate curves becomes more pronounced with larger target spike counts (Figure 3E and 7E). The striking differences between different ganglion cells in the nonlinearities of signal integration raise the question of the associated visual functions. To illustrate the effects of the observed Adenosine receptive field nonlinearities, let us therefore consider a simple visual stimulus, which contains a large dim object as well as a group of several small objects at high contrast (Figure 8A). When viewed through linear receptive fields, both the large dim object and the area with the small high-contrast objects appear equally prominent (Figure 8C). Receptive fields that integrate their subunits with a threshold-quadratic nonlinearity, however, emphasize the high-contrast region (Figure 8D), whereas nonlinear integration in the fashion of homogeneity detectors facilitates the detection of the large dim object while being insensitive to high-contrast clutter (Figure 8E). This suggests that homogeneity detectors contribute particularly to the detection of large objects.