​Attentional switching of connectivity between visual and memory systems.

Attention amplifies neural responses within areas coding for goal-relevant information and also strengthens the coupling between these areas. This modulation has been found repeatedly in ventral visual cortex and linked to the behavioral effects of attention on perception. However, attention also has a powerful effect on learning and memory behavior, suggesting that such modulation may impact the medial temporal lobe (MTL) memory system. Here we investigated this possibility by examining how visual input into the MTL gets prioritized based on top-down attentional goals. In particular, we focused on two cortical structures, perirhinal cortex (PRC) and parahippocampal cortex (PHC), which project to the entorhinal cortex and hippocampus. These regions are thought to belong to different pathways, with PRC involved in face and object processing and PHC involved in scene and spatial processing. However, in addition to these static networks, we hypothesized that these regions would dynamically couple with ventral visual cortex depending on whether attention is directed to their preferred content - that is, visual cortex would be more coupled with PRC when attending to faces and with PHC when attending to scenes. To test this prediction, subjects were exposed to blocks of composite images containing both face and scene information, but attended to only one category. We measured background connectivity between PRC, PHC, and ventral visual cortex separately during the face and scene attentional states. Region-of-interest and voxelwise analyses revealed an interaction consistent with our prediction, with visual cortex coupling relatively more with PRC during face attention and PHC during scene attention. Interestingly, the degree to which attention modulated a voxel's connectivity was negatively correlated with the voxel's category selectivity - voxels with mixed selectivity had the most malleable connectivity. These findings suggest that attention determines which MTL structures receive the most input from ventral visual cortex. Meeting abstract presented at VSS 2015.