Abstract
Neuroimaging studies have identified the human superior temporal sulcus (STSp) to have brain responses correlated to the perception of biological motion (e.g. Grossman et al., 2000; Allison, Puce & McCarthy, 2000). The human STS is believed to be the homologue to monkey superior temporal polysensory area (STPa), in which single-unit physiology studies have shown neurons responsive to biological motion (e.g. Perrett et al., 1996). Many neurons in monkey STPa are reported as sharply tuned to particular actions, and are proposed to form the basis of action recognition. The aim of this study is to determine whether the human STS, like the monkey STPa, has unique neural populations supporting the recognition of different biological actions. If such neuronal populations exist, they are likely organized at a sub-voxel spatial scale. To overcome these spatial resolution limitations of functional magnetic resonance imaging, we have measured the information content in the fMRI BOLD responses using support vector machines in conjunction with multi-voxel pattern analysis. We specifically measured whether the STS response discriminates between different biological action as well as between those actions and motion-matched non-biological control stimuli (‘scrambled’ motion). Human subjects viewed blocks of three different conditions (jumping jack, profile view of walker and hand waving in the air), and a scrambled animation condition. We measured classification performance in the STSp and in the motion sensitive hMT+, both independently localized in separate scans. We found above chance classification performance in these regions, which is evidence of sufficient information in the BOLD pattern to discriminate different unique actions. This study provides insight into neural activity at the sub-voxel level in human brain areas involved in biological motion perception, and our findings suggest that action recognition is supported by highly-tuned neuronal ensembles in visual cortex.