Cross-modal decoding of emotional expressions in fMRI-Cross-session and cross-sample replication

The theory of embodied simulation suggests a common neuronal representation for action and perception in mirror neurons (MN) that allows an automatic understanding of another person's mental state. Multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) data enables a joint investigation of the MN properties cross-modality and action specificity with high spatial sensitivity. In repeated-measures and independent samples, we measured BOLD-fMRI activation during a social-cognitive paradigm, which included the imitation, execution, and observation of a facial expression of fear or anger. Using support vector machines in a region of interest and a searchlight-based within-subject approach, we classified the emotional content first within modalities and subsequently across modalities. Of main interest were regions of the MN and the emotional face processing system. A two-step permutation scheme served to evaluate significance of classification accuracies. Additionally, we analyzed cross-session and cross-sample replicability. Classification of emotional content was significantly above chance within-modality in the execution and imitation condition with replication across sessions and across samples, but not in the observation condition. Cross-modal classification was possible when trained on the execution condition and tested on the imitation condition with cross-session replication. The searchlight analysis revealed additional areas exhibiting action specificity and cross-modality, mainly in the prefrontal cortex. We demonstrate replicability of brain regions with action specific and cross-modal representations of fear and anger for execution and imitation. Since we could not find a shared neural representation of emotions within the observation modality, our results only partially lend support to the embodied simulation theory. We conclude that activation in MN regions is less robust and less clearly distinguishable during observation than motor tasks.