Abstract
Early functional magnetic resonance imaging (fMRI) studies revealed a constellation of areas in ventral occipital and temporal cortices (VOT) that respond selectively to objects, and areas that respond selectively to specific categories of objects (e.g. faces). Using pattern analysis and a large number of object exemplars, fMRI studies later observed object-category information to be distributed across VOT. More recently, studies have found that the brain maintains a representational space of objects, in which conventional categories form clusters and individual exemplars can be differentiated. In the present study, we used magnetoencephalography (MEG) and pattern analysis to examine how this representational space emerges over time. Participants viewed exemplars of six categories of objects (human faces, animal faces, human bodies, animal bodies, natural objects, and man-made objects) while brain activity was recorded using MEG. Using a temporal sliding window, we studied the pattern similarity of the stimuli in the evoked response. The analysis revealed a dynamic representation that evolves over time. Shortly after stimulus onset, the evoked patterns for exemplars became distinct for individual exemplars with no apparent coherence among categories. Human faces then emerged as a distinct category. The geometry of the space then evolved to a state that exhibited a predominant structure distinguishing between animate and inanimate objects. Shortly thereafter, substructure emerged within this structure that distinguished animate object categories (human faces, animal faces, human bodies, animal bodies). The outcome of the study shows that the brain’s representation of objects dynamically changes over time, presumably reflecting distinct stages in the visual processing of objects.
Meeting abstract presented at VSS 2012