Mapping invariance in the ventral occipital temporal pathway

• Main Author: Racey, Chris
• Other Authors: Hartley, Tom
• Published: University of York 2012
• Subjects: 612.8
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.568123

This thesis is concerned with the organisation of the human ventral occipital temporal pathway, part of the brain involved in the recognition of objects and processing of scenes. The architecture of this system has been extensively investigated with functional magnetic resonance imaging (fMRI) which reveals the engagement of parts of the brain in different tasks. Conventional analyses have been used to identify discrete, circumscribed brain regions which appear to be specialised for processing particular categories of stimuli, such as faces or landscapes. In characterising the function of these regions, an important factor in previous research has been to establish whether or not they are invariant to more basic sensory properties such as size or 3D orientation, which change as places and objects are encountered from different points of view. A more recent suggestion is that the ventral occipital temporal pathway, rather than being comprised of discrete regions, is better conceived of as a continuous topographical map, or multiple overlapping maps. Within this framework, fundamental computational requirements, such as the need to establish different forms of invariance, may provide overarching organising principles governing the spatial arrangement of processing within ventral visual cortex. This more nuanced view of the functional architecture of the visual system demands a more nuanced view of invariance. Through a series of experiments using fMRI in conjunction with a parametric adaptation paradigm, this thesis investigates the invariance of neural representations as a continuous variable that can be explicitly quantified rather than an all-or-none phenomenon. This novel approach reveals structure and symmetry normally hidden in conventional analyses. Invariance is found to vary by degree throughout the ventral surface of the brain depending on the nature of the stimulus and the types of processing required. Different forms of invariance can be dissociated from one another in terms of distinct spatial patterns of parametric adaptation that different forms of visual change generate.