How low-level segregation within the visual system impacts high-level object recognition

A long tradition in vision research points to the existence of multiple visual pathways that process different features of the incoming input. This segregation begins at the retina, where different types of ganglion cells can be distinguished by their patterns of dendritic arborization, and the types of visual stimuli that are most effective in exciting them. Those distinctions are carried forward to subcortical and cortical levels as different channels within the visual system. The most salient distinction is between magnocellular (M) and parvocellular (P) channels within the visual system, which receive their input from parasol and midget ganglion cells, respectively. The existence of different channels within the visual system places strong constraints on the types of information to which we are sensitive in our environment, both during online processing as well as on an evolutionary timescale. The central question that we want to address is whether the signatures of low-level segregation within the visual system affect the organization of high-level object representations. The novel insight we explore is that processing of different categories or classes of objects differentially depends on (or differentially weights) input from different low-level visual pathways. To test this hypothesis, we use fMRI as an index of neural activity and psychophysically titrated images of real objects as stimuli. This approach of seeking to understand high-level object recognition through the lens of low-level visual constraints is largely unexplored, but we believe it to be an incredibly fertile enterprise for understanding the neural constraints that shape object knowledge.