Thus, one advantage of a subcortical demodulating nonlinearity is that it simplifies the construction of a form-cue invariant circuit. Like many other nonlinear scene representations, the neural representation of ICs has been thought to originate in cortex (Baker and Mareschal, 2001, Song and Baker, 2007 and von der Heydt and Peterhans, 1989). However, since theoretical work has shown that a demodulating nonlinearity will detect ICs (Daugman and Downing, 1995), we hypothesized that AZD6244 purchase a neural representation of ICs
may originate subcortically with Y cells. To examine this possibility, we recorded the responses of a small number of LGN Y cells to abutting grating stimuli used to study cortical processing of ICs (Grosof et al., 1993 and Song and Baker, 2007). Y cell responses invariably oscillated at the frequency of ICs/s, indicating that the ICs were detected (Figure S4). This suggests that by demodulating visual signals, Y cells may encode a variety of complex image features whose LDK378 detection was previously thought to require cortical processing. Neural responses to high spatiotemporal frequencies are significantly attenuated between the LGN and primary visual cortex (Derrington and Fuchs, 1979, Hawken et al., 1996, Ikeda and Wright, 1975 and Movshon et al., 1978).
This lowpass filtering in the geniculocortical transformation is thought to limit the perception of dynamic visual scenes (Hawken et al., 1996 and Zhang et al., 2007). For example, the imperceptible flicker of Florfenicol 60 Hz
monitor refresh drives many subcortical but few cortical neurons (Wollman and Palmer, 1995). However, by extracting envelope TFs subcortically, high spatiotemporal frequencies that are filtered out in the geniculocortical transformation may still influence perception. Consider the invariant carrier TF tuning of the cell shown in Figure 2E. Whether the component TFs are low or high, the signal transmitted to cortex is indistinguishable (it oscillates at the envelope TF). Because only low envelope TFs are represented by Y cells (Table 1; Figure S5B), they are not filtered out in the geniculocortical transformation. As such, a cortical neuron innervated by this cell should have also responded to the motion of the envelope without regard to the carrier TF. Other Y cells, like the one in Figure 2D, only project information about interference patterns to cortex when the component frequencies are so high that it is unlikely that the envelope TF can be computed in cortex (i.e., from the output of area 17). This implies that image components whose spatiotemporal frequencies are too high to pass the geniculocortical filter can still drive cortical responses, and as a result, likely influence perception.