Abstract
Neuroimaging studies have shown that viewing animate objects like people and animals evokes stronger activity in posterior superior temporal sulcus (pSTS) and lateral fusiform gyrus than does viewing inanimate objects, whereas viewing inanimate objects evokes stronger activity in medial fusiform and lingual gyri and inferior lateral temporal lobes. Less is known about representational structure within the domain of animate categories, for example how different animals are represented. Using fMRI we recorded brain activity (N=11) associated with viewing 12 animal species -- 4 each from mammals, reptiles, and bugs. Using a novel technique for identifying shared representational structure that involves clustering searchlight-defined dissimilarity matrices, we defined regions of interest that included lateral and ventral occipito-temporal cortex, which we refer to as lateral occipital complex (LOC). We used multivariate techniques including pattern classification and similarity structure analysis to explore representation within this region. Classification accuracies were highly significant across subjects for within and between class discriminations. Similarity structures in LOC were highly reproducible across subjects with average correlation between subjects of r = .81. Multidimensional scaling revealed a dimension spanning from mammals to reptiles to bugs as a common axis across subjects. The projection of the primary MDS dimension onto the beta weights for the animal categories reveals greater activity for mammals than for bugs in lateral fusiform and pSTS, and greater activity for bugs than for mammals in medial fusiform and lateral inferior temporal cortex. The mammals, which are both subjectively and biologically closer to humans, produced activity similar to that associated with viewing animate objects, while bugs, which are subjectively and biologically distant from humans produced patterns similar to those for inanimate objects. These findings suggest that animal categories fall along a continuum in representational space that is predictable by the degree of animacy exhibited by each category.
Meeting abstract presented at VSS 2012