We rely for object recognition on hierarchical processing of visual input by the ventral stream of cortical visual areas. The chain begins in low-order cortex (including area V2) where neuronal activity represents local features and terminates in inferotemporal cortex (including area TE) where neuronal activity represents global images. Neuronal function in the ventral stream is subject to alteration by visual experience. One particularly robust form of visual plasticity is familiarity suppression. Familiarity suppression can be induced in TE by allowing subjects to view the same set of images repeatedly over the course of days and weeks. At the end of the familiarization period, the mean strength of neuronal responses in TE is less for the familiar images than for novel controls. On the assumption that familiarity suppression is specific to TE, it has been thought to serve some function related to late-stage processing of the global image. Recently, however, familiarity suppression has been shown to occur even in V2, where neurons represent local features and not entire objects. On the basis of this observation, we propose that familiarity suppression, far from being specific to late-stage processing of the global image, arises from principles of unsupervised statistical learning operative at each stage of the ventral stream and expressed at each stage relative to the nature and scale of represented visual information. We will test this general idea by using semi-chronic multielectrode arrays to monitor neuronal population activity simultaneously in V2 and TE during image familiarization. Experiments under Aim 1 will test the hypothesis that familiarity suppression develops independently in V2 and is not just fed back from TE. Experiments under Aim 2 will test the hypothesis that familiarization enhances the cooperative representation of an image by neuronal populations in V2 and TE. Experiments under Aim 3 will test the idea that the synaptic alterations underlying the impact of familiarization obey a Hebbian learning rule.
The proposed experiments will improve our knowledge about the neural mechanisms that underlie object vision and in particular the changes and improvements that occur as images become familiar. Better understanding of these issues will improve our ability to diagnose and ameliorate the grave disorders of vision that follow from injury to occipital and temporal lobe cortex in humans.
