In the 1980's Antonio Damasio proposed a neural architecture in which neurons in high level association cortex serve as "converge-divergence zones", or CDZs. These neurons receive bottom-up inputs from multiple sensory modalities (convergence), and can therefore be activated by both the sight and sound of an object. However, they also send reciprocal projections back to lower level sensory cortices (divergence). Through these top-down connections CDZs are able to re-instantiate representations in early sensory cortices.

One prediction of this framework is that bottom-up activation triggered by a sensory stimulus in one modality will always be accompanied by associated top-down activation of other sensory modalities. For example, the sight of a bell would ultimately lead to activation of associated representations in auditory cortex as we hear the bell in our "mind's ear". Using multi-variate pattern analysis (MVPA), we found evidence that this is indeed the case (Meyer et al., 2010). In this first study, subjects watched a series of video clips that were silent, but implied sounds (e.g. a rooster crowing). Using voxels anatomically restricted to auditory cortex, we found that a support vector machine (SVM) could predict which video clip the subject was watching significantly better than chance. Thus the videos in our experiment evoked content-specific activation patterns in auditory cortex.

We next extended this finding into the somatosensory domain (Meyer et al., 2011). In that study subjects watched five videos that showed a hand touching one of several textured objects. Our SVM was able to predict which touch-implying video the subject saw using only voxels from somatosensory cortex. Furthermore, we found that these patterns in somatosensory cortex were similar enough across individuals that we could predict which stimulus someone was viewing using a classifier trained on data from different individuals (Kaplan et al., 2012).

Our current work seeks to extend these findings by directly examining interactions between stimuli presented through multiple sensory modalities.

This line of work focuses on understanding the neural systems involved in identity and self concept, including studies of how we recognize our own faces and voices, how we engage in the mental process of of self-reflection, and how we form a self-narrative or autobiographical self. I am also interested in how social group membership affects self-related neural processes.

The goal of this project is to understand the neural mechanisms that underlie belief. We have found that specific brain systems are activated when people believe compared with when they disbelieve, and that these systems appear to be engaged similarly regardless of the content of those beliefs (religious or nonreligious). We have also explored the process of resistance to belief change for political beliefs.

This work focuses on the neural basis of social cognition, including how we understand and empathize with others, and how our social relationships affect our brain functioning.

Morteza Dehghani and I have started the Brain Imaging and Machine Learning journal club at USC to bring together people on campus interested in these topics. If you're interested in joining us, you can visit the wiki or to be added to the mailing list.