FMRI approaches to high-level vision: object representation and recognition
Functional magnetic resonance imaging in humans has identified subregions in ventral visual cortex selectively involved in processing of object shape. Some of these respond similarly to different objects, others more specifically to certain categories as faces or scenes. However, the detailed level of representation of objects in these areas is still not well understood. FMRI examinations adressing questions of object representation have been complicated by the fact that at the neural level, individual objects are probably represented in a way that is beyond the spatial resolution of conventional fMRI. One way to achieve “hyper-resolution” has been to use repetition paradigms that make indirect inferences about object representation through changes in activation following repeated presentation. I will present parts of my work using these methods to study the representation of faces in ventral visual cortex, and also address limitations of this approach. Next, I will present data showing that similar questions of object representation can be studied in comparisons of direct evoked activity when applying multivariate pattern recognition to fMRI data. These methods allowed to detect information about object category and exemplar in cortical areas that are commonly considered a general shape processing system (unselective for different object types), and open up new ways to investigate neural correlates of object constancy in human cortex. The last part of the talk will focus on the role of ventral visual cortex in explicit recognition of objects, and its top-down modulation by higher order areas in the presence of prior knowledge about object identity, presenting fMRI data and analyses of interregional coupling.