1B) are characterized by positive responses for both directions of the grating reversals for several grating positions, in particular when positive and negative contrast are balanced over the receptive field. These response characteristics cannot be explained by a model with linear integration of light signals over space. More formally, the distinction between linear X cells and nonlinear Y cells is often based on computing the amplitudes of the first

and the second harmonic of the firing rate in response to the periodic grating reversals (Hochstein and Shapley, 1976). X cell responses are dominated by the first harmonic (Fig. 1C), whereas the fact that Y cells can respond to both grating reversals leads to frequency doubling and an often dominant second harmonic in the firing rate profile (Fig. 1D). Note that the linear spatial integration in X cells does not imply that these cells respond to the two opposite grating reversals with firing rate profiles that are ATM Kinase Inhibitor price equal in magnitude with opposite signs, as would be expected for a completely linear system. In fact, retinal ganglion cells, like most other neurons in the nervous system, display a nonlinear dependence of the firing rate on stimulus strength simply because the spiking itself is subject to a threshold and potentially saturation. Thus, positive responses upon grating reversals are typically more pronounced than the amount of suppression observed for

the opposing reversal. This can www.selleckchem.com/JNK.html be viewed as a nonlinear transformation of the integrated activation signal. This nonlinearity, however, does not affect how signals are integrated over space prior to this output transformation. We will return to this distinction between different nonlinear stages in the stimulus–response relation of ganglion cells below. The separation between X cells and Y cells does

not always appear clear-cut and may in some systems rather represent the extremes of a continuum with different degrees of nonlinear integration, as reported, for example, for mouse retina (Carcieri et al., 2003). Moreover, and the fact that anatomical investigations typically distinguish around ten to twenty different types of ganglion cells (Masland, 2001, Rockhill et al., 2002, Dacey, 2004, Kong et al., 2005, Coombs et al., 2006, Field and Chichilnisky, 2007 and Masland, 2012) suggests that the classification of X and Y cells represents only a coarse categorization, which might allow further division into subtypes, for example, by refined measurements of the spatial integration characteristics. The finding of nonlinearly integrating ganglion cells has led to the development of subfield models, which describe the receptive field structure of Y cells as composed of spatial subfields whose signals are nonlinearly combined (Fig. 2). These model efforts were initiated by measurements of Y cell responses to sinusoidal temporal modulations of different spatial patterns (Hochstein and Shapley, 1976).