Abstract
Where do you know who you know? There are faces that are instantly recognizable and nameable. Conversely, there are names that automatically bring up an associated face. Some neurons are likely to respond to both the name and the face of these familiar individuals – in fact, such neurons with an invariant, explicit and selective response to both famous and familiar individuals, e.g. Jennifer Aniston, are routinely found with the aid of implanted microelectrodes in the medial temporal lobes of human patients (Quian Quiroga, Reddy, Kreiman, Koch, & Fried, 2005). We set out to investigate such modality-independent, conceptual representations with fMRI. We conducted an event-related fMRI study in which subjects were asked to recognize and name three famous actors (Brad Pitt, Matt Damon and Tom Cruise) from various pictures or to read their written name, on separate trials within each run. We used high resolution fMRI (2x2x2mm voxels) with 30 axial slices covering the occipital and temporal lobes. Performing multivariate pattern analysis (linear support vector machine classification, with leave-one-run-out cross validation) within a searchlight volume operating throughout the brain for each subject, we demonstrated at the group level (seven subjects) that activity patterns specific to the name of a given actor allow decoding of their identity upon presentation of their face, and vice-versa. Such patterns were only found in the left anterior temporal lobe, consistent with a large body of lesion and imaging studies implicating this region in the retrieval of proper names (Patterson, Nestor, & Rogers, 2007). This result opens up a new avenue of research which has the potential to address critical issues in the field of semantic memory and concept retrieval in the human brain.
Meeting abstract presented at VSS 2012