Abstract
The specialized neural functions associated with the visual processing of faces and bodies span multiple cortical regions in ventral temporal (VT) and superior temporal (ST) cortex. We used pattern classification to separate the neural activity elicited in response to viewing moving and static presentations of faces and bodies. Using fMRI, participants (n = 8) viewed 12-second blocks of videos of people walking toward a camera and 12-second sequences of the "best" static images from the videos. For dynamic and static presentation types, participants saw: 1.) videos/images of the whole person (WP); 2.) videos/images of the body with the face pixilated (B); and 3.) videos/images of the face with the body obscured (F). Pattern classifiers were implemented to discriminate all pairs of conditions (e.g., dynamic F vs. dynamic B) in each participant’s brain. Neural activation patterns were highly discriminable for static-to-static (Mean d’ = 2.22, se = .21) and motion-to-motion (Mean d’ = 2.15, se =.24) comparisons (F vs. B, F vs. WP, B vs. WP) in functionally localized face/object selective regions of VT cortex. For both cases, F vs. B and F vs. WP activations were more separable than B vs. WP. In ST cortex, these conditions were separable also, but with the motion-to-motion F vs. B comparison significantly more discriminable than the other comparisons. For motion vs. static conditions, neural activity maps in the VT cortex were separable only for cross-condition comparisons that included the face (F vs. B, F vs. WP; Mean d’ = 1.58, se = .35). Within a condition (e.g., moving F vs. static F), the neural activation maps could not be separated at levels above chance (Mean d’ = -0.23, se =.48). This study offers a systematic dissection of the neural representations of human face and body motion in a natural stimulus.
Meeting abstract presented at VSS 2012