| Faces are a rich source of social information. Facial expressions can convey information about emotional states, and neutral expressions are believed by many human observers to carry information about personality traits, although the accuracy of this information is low. In this thesis, I use fMRI to investigate how the brain responds to facial expressions and face traits. In Experiment 1, I use multivoxel pattern analysis to show that the posterior superior temporal sulcus contains distributed representations of facial expressions. I further show that the similarity structure of these representations correlates with the perceptual similarity structure of the expressions. In Experiment 2, I turn my attention to the emotion overgeneralization theory of trait inferences. Neutral faces can be rated on many traits, such as trustworthiness, aggression, and competence. Using applications from computer vision, I show that neutral faces perceived to possess various personality traits contain objective resemblance to emotional expressions. These results support the idea that trait inferences about neutral faces are in part the result of an overgeneralization of emotion recognition systems. In Experiments 3--5, I investigate the neural responses to faces varying on perceived trustworthiness and dominance. These two dimensions have been previously shown to explain most of the variance in face trait ratings. In Experiments 3 and 4, I show that the fusiform gyrus, lateral occipital lobe, lateral prefrontal cortex, and portions of the parietal lobe all showed linear responses to face trustworthiness, with greater responses to untrustworthy than trustworthy faces. Additionally, there were extensive quadratic responses to face trustworthiness in the fusiform gyrus and dorsal amygdala, with greater responses to the extremes of face trustworthiness than to faces in the middle of the continuum. Responses to face dominance were generally less extensive, but included a quadratic response in the fusiform gyrus. The quadratic response to these social dimensions, while extensive throughout much of the brain, may not be specific to social dimensions. In Experiment 5, I show that the quadratic response to face trustworthiness in the amygdala and fusiform area is driven at least in part by the general distance from the mean of face space. |
